Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber

https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-6-flash-3-5-flash-lite-3-5-flash-cyber/

by logickkk1 • 1 day ago

739 547 1 day ago

blog.google

https://console.cloud.google.com/agent-platform/publishers/g...

yanis_t• 1 day ago

The benchmarks are not particularly impressive. I suppose they needed to release something since the long pause. But not clear why would I use it now.

dumberquestions3• 1 day ago

"..and in some benchmarks like DeepSWE by Datacurve, we observe up to 65%, all at a lower cost per output token."

"3.6 Flash delivers higher precision with fewer unwanted code edits and reduced execution loops, as seen in DeepSWE (49% vs. 37%)"

So which one is it? 65% or 49%?

petu• 1 day ago

First sentence is about token efficiency.

dumberquestions3• 1 day ago

You're right, should've gotten some LLM to summarize it instead of skimming.

jgbuddy2• 1 day ago

It is both less intelligent and more expensive than GLM-5.2, while being closed weight.

drob5183• 1 day ago

But they make up for it by shipping it late.

SwellJoe2• 1 day ago

It's also got vision and audio. So, the better comparison is any of the other large Chinese open models that are better and cheaper than Gemini Flash.

lenerdenator2• 1 day ago

It'd be interesting to know how much the Intelligence as a Service angle serves as a value-add in the minds of Google's executives.

You can get decent open-weight models now. That's not difficult. The difficulty is 1) running them and 2) compliance.

My company runs Claude on GCP's Vertex AI solution. We're in the US healthcare IT space, so the models need to be from somewhere that American healthcare agencies and companies have traditionally been okay with sourcing code from - which means the US, Canada, and maybe Europe. The stuff that handles PHI/PII must be in the US. The expense of hosting is more of a PITA than most customers want to go through this early in the technology's lifecycle, and intelligence gains are simply a matter of degree for most business tasks.

In theory, we could find some open-weight model (likely from China) for our development agentic work and host it anywhere you can host AI models. We don't, though, and I think Google, OpenAI/Microsoft, and Anthropic see that as the core of their business.

dyauspitr2• 1 day ago

It’s multimodal though.

ur-whale5• 1 day ago

It’s multimodal though.

Sure.

And how does that make your day better? I know it does not improve my work in any way shape or form.

I'll take a better coding model that's not multi-modal any time.

If I need an LLM to do images or sound, I'd rather use a dedicated one instead of a jack-of-all-trades-master-of-none model.

JacobAsmuth6• 1 day ago

It's also about 15x faster.

SubiculumCode2• 1 day ago

Is that statement based on token price? More and more it seems that $/token hides as much as it reveals. Token efficiency, tokenizer differences, etc. I'm not saying that you are wrong, I am just saying it is becoming a bit more difficult making statements like this without a bit more research.

jgbuddy2• 1 day ago

It's based on the Artificial Analysis "Intelligence Index vs. Cost per Intelligence Index Task" here:

https://artificialanalysis.ai/#intelligence-comparison-tabs

Differences in token "density" are accounted for by pricing per task

jbellis• 1 day ago

It's roughly equal on price and intelligence as GLM 5.2 while being ~8x faster.

kzrdude6• 1 day ago

Google's models are always very well spoken and much more pleasant to talk to. And I've used the open weight models a lot, too.

metalliqaz• 1 day ago

Other discussion from a few minutes earlier: https://news.ycombinator.com/item?id=48993130

velominati• 1 day ago

Wow - Google does not even bother to show benchmarks of these models compared to the frontier and Chinese labs - only against previous versions. I'm not surprised. Having worked there for years it was amazing just how inwardly looking the company is.

dvduval• 1 day ago

It does seem like their releases are getting closer together. I get the feeling they realized they were trying to roll out to their entire ecosystem and now they’re focusing more just directly on the AI model itself. I think give it a little time and they’ll start to be one of the competitors too.

singingtoday• 1 day ago

I'm more excited for 3.5 pro. Gemini has fallen behind in some areas, but is still one of the best multimodal models.

Has anybody found any models better at image or audio analysis?

ianhawes• 1 day ago

Came here to ask basically this. We use 3.1 Pro internally and it's great.

npn2• 1 day ago

tested the models on aistudio. despite that the knowledge cut off is march 2026 it still knows nothing about 2025!

you can check by asking "list notable world events in 2025, only list unplanned" on aistudio. or you can ask for Charlie Kirk, it also does not know. I tried it multiple time to ensure that I didn't not get routed to older models!

but google has search

irrelevant, without deeper knowledge about cutting edge technologies or latest libraries, all of it suggestions are crap. even you ask it to search it will still use outdated keyword thus only getting outdated information.

in other word, what a disaster!

npn2• 14 hours ago

update: the knowledge cut off date is "unknown" now.

funny because some people downvoted me believed that there is no relation between knowledge cut off date and real world events. that's not how it works!

nsbk• 1 day ago

It is 17% more token-efficient than 3.5 and performs significantly better in coding and tool usage benchmarks.

It is also cheaper than 3.5:

This enhanced efficiency is also combined with a lower price than 3.5 Flash. At $1.50/1M input tokens and $7.50/1M output tokens, 3.6 Flash reduces the overall cost per agentic task, making agents more cost-effective to build and run.

ilreb• 1 day ago

dupe? https://news.ycombinator.com/item?id=48993130

m_w_2• 1 day ago

It's a bit disheartening to see no comparison to other models here - and I'm not sure this pushes the curve anywhere. 3.6 flash is more expensive than GLM 5.2 - but seemingly worse, although this post is really light (lite?) on details.

It seemed for a time that Google had finally gotten the ball rolling, but I'm doubting that more and more as time passes. We'll see what happens with 3.5 pro I suppose.

dd8601fn2• 1 day ago

really light (lite?) on

Light. Lite is product marketing seepage.

crab_galaxy• 1 day ago

Yeah that’s the joke :p

armarr• 1 day ago

GLM was twice as verbose running the Artificial Analysis benchmark. So it ends up being more expensive

maxloh2• 1 day ago

Not really. Gemini 3.6 Flash actually cost $0.01 more per task, compared to GLM 5.2.

https://artificialanalysis.ai/models/gemini-3-6-flash

Havoc3• 1 day ago

verbose

GLM defaults to max effort btw

https://docs.together.ai/docs/glm-5.2-quickstart#reasoning-e...

XCSme8• 1 day ago

Here, my comparison of 3.6 Flash vs Sol vs Luna vs Terra: https://aibenchy.com/compare/google-gemini-3-6-flash-medium/...

jdthedisciple3• 1 day ago

How does your comparison work? It places Gemini 3.6 Flash Medium above GPT 5.6 Sol High and Fable 5 Medium, which makes me skeptical because that... would be making headlines that I'm not seeing right now.

dudeinhawaii2• 16 hours ago

You should really provide more on your methodology because as it stands, it really doesn't pass the sniff test. GPT-5.6 Sol on Low beats Fable Medium by 10% and Gemini-3.6 Flash then beats them both? Fable is number 20?

This does not match any lived experience or developer experience.

It'd be helpful to know _what_ you're testing and break that out by dimension. You mention randomly selected questions. How does that work?

With n=22 and binary pass/fail, the 95% confidence interval on a pass rate spans roughly (+-)15-20 percentage points. There's just not enough data ironically, for this leaderboard to mean anything. Ranks #5 through #25 are statistically indistinguishable

CSMastermind2• 1 day ago

All the benchmarks I see put it around the capabilities of Opus 4.8 Medium or Sonnet 5 High.

As far as I can tell it's slightly better than GLM 5.2.

zwaps4• 1 day ago

according to AA it's not better than GLM 5.2 and that's surprising to me

sczi• 1 day ago

The one thing I've found google's models to be the best at is proofreading text in non-english languages. Probably because I imagine they have the most training data for it as Google probably has the most complete archive of the internet.

florakel• 1 day ago

It’s really surprising. When Apple announced the multi-billion dollar deal with Google to power Apple Intelligence I thought great things were coming. Instead we are getting more and more bad news: delayed Pro models and AI leadership leaving. I wonder if Apple know something the rest of us don’t know or if they are already regretting their decision.

revolvingthrow2• 1 day ago

What Apple wants out of Google is Siri that runs at 8gb ram and isn’t a horrible embarrassment that feels like a primitive markov chain. Given how good Gemma 4 is, Google can squeeze some serious performance in small models. Whether they can make bleeding edge models is irrelevant to Apple.

Petersipoi2• 1 day ago

As someone on the Apple beta.. the model is almost completely irrelevant to the experience. Apple has gone and done Apple things by nerfing the experience so completely that almost any model in the past year would be fine. I still reach for ChatGPT/Claude/Grok constantly instead of the AI toy that Apple calls the new Siri.

WarmWash10• 1 day ago

Besides Apple apparently making Siri AI model agnostic, the choice to go with Google was almost certainly for practical reasons. Google is a low-risk established player that already has a long work history with Apple. Google also isn't in an existential battle to establish themselves, Gemini still amounts to just another project at Google. There is tangible non-zero risk that either OAI or Anthropic will be gone in 5 years, or will be forced to leave Apple high and dry to save themselves. There is almost no risk Google will be in either such position. And worst case scenario, Google has incredibly deep pockets should Apple pursue a "refund."

winstonp• 1 day ago

Apple isn't counting on their model to be a frontier coding and cowork model. Gemini is perfectly fine for the tasks that new Siri is supposed to be doing.

Oras• 19 hours ago

The bar was quite low with Siri that adding any model would make a huge difference.

ur-whale5• 1 day ago

It's a bit disheartening to see no comparison to other models here

Disheartening, but not surprising: the comparison would not be very flattering for Google.

Computer0• 24 hours ago

My usage of GLM 5.2 has been defined by slow throughput and flaky providers, in many (definitely not all!) applications a dumber, faster, and more consistent model makes more sense to me.

b473a2• 1 day ago

No word about updating Jules, which is still stuck on 3.1 Pro. I get that it's probably niche but I've really appreciated basically being able to give directions to Jules on my phone, then reviewing and merging a GitHub PR fifteen minutes later. It's been great for getting some progress in on a few personal projects during my commute when I can't exactly pull out my laptop.

Anyone have any good alternatives?

haberdasher• 1 day ago

Claude Code

steven_pareto• 1 day ago

If you own a Raspberry Pi or similar: Hermes + Tailscale + iSH over tmux.

aweb• 1 day ago

Both Claude and Codex can code in the cloud, it works quite well!

I tested Jules and while the idea is good in theory, I found the model's intelligence to be very lackluster.

b473a2• 1 day ago

Shame. I'm on the $20/mo Gemini Pro plan because the 5tb of cloud storage and the youtube premium lite were good enough perks, and my coding complexity needs were light enough for me to overlook Claude or Codex. But Antigravity is working better than Jules and it's basically giving me a taste of what I'm missing and it's harder to justify not trying out the competitors.

christoff12• 1 day ago

I have no affiliations with the team or product, but Superconductor reminded me of Jules when I tried it a couple of months ago.

It might be overkill features-wise, but there's a free tier and it likely won't be left for dead anytime soon.

bespokedevelopr• 1 day ago

I do not, however I am curious about Jules support. I didn't know if this was a dead project or not. Seemed really interesting but then I didn't see much development/announcements/discussions around it. Last update from their changelog was as you said 3.1-pro support in March.

geooff_• 1 day ago

At this point just put the Pareto in the bag bruh

ConfusedDog• 1 day ago

Why would 3.6 flash perform a little worse than 3.5 flash on Artificial Analysis Coding Index...

https://artificialanalysis.ai/models/gemini-3-6-flash?intell...

sosodev2• 1 day ago

Because AA Coding "Index" consists only of two benchmarks (Terminal-Bench v2.1, SciCode) and generally fails to be meaningfully representative of agentic coding capabilities.

Alifatisk3• 1 day ago

Whats a better option for AA Coding Index?

firethunder7• 1 day ago

AA coding index has been updated to use DeepSWE, Terminal-Bench v2, and SWE-Atlas-QnA.

wmedrano• 1 day ago

Could be a good tradeoff for the flash model though. 3.5 -> 3.6 is a tiny bit cheaper and maybe faster?

artificialanalysis.ai has it going from 165 tps -> 304 tps. openrouter.ai needs more data but it has it going from ~100 tps -> ~150 tps, though at peak 3.5 has reached 156tps.

dankai• 1 day ago

Unfortunately says more about how competitive 3.5 pro would be today at the frontier if they forgo it for 3.6 flash.

catigula• 1 day ago

"We made 3.6/4 Pro, but it sucks, so this is the distilled model" vibes.

drob5183• 1 day ago

That’s the fear.

mfkrause• 1 day ago

Pretty underwhelming, as expected honestly. I don't want to know what morale is like at DeepMind right now.

drob5183• 1 day ago

Yep, agreed. They still are not releasing anything frontier-class (Gemini Pro) at this point. Feels to me that they keep getting scooped by others (e.g. Kimi 3) and then are retrenching.

WarmWash10• 1 day ago

Especially when Google owns 15% of anthropic and serves them compute. Double especially when your boss (Hassibis) is also an early investor in Anthropic. Hell his NW might be more Anthropic than Google.

kilroy123• 1 day ago

I deeply wish Google would focus on models like Gemma. Small, powerful, open-weight models you can run on phones or regular computer hardware.

lanthissa2• 1 day ago

i mean eventually then will, losing means open source, vertically integrated hardware means you can opensource and win on cost

mediaman4• 1 day ago

Gemma 4 was released in April. It's a good series of multimodal models.

accountrequired2• 1 day ago

gemma 4 thinks joe biden is president

doctoboggan• 1 day ago

I have a side business selling custom fingerprint jewelry and I use gemini nano banana to clean up customer submitted fingerprint images. This was a step I used to do by hand at 10 - 15 minutes per image and nano banana is the first model that is able to do the task (it is astonishingly good at it). I can't wait to see what the next nano banana can do, hopefully its released soon.

cube00• 1 day ago

Are your customers clearly informed that you're sending their immutable fingerprints to an AI service?

poisonborz• 1 day ago

Yes this is extremely unresponsible if so. Fingerprints are legally protected biometric data in most juristictions.

dinkelberg• 21 hours ago

Their shop is linked to in the bio. They don't seem to have a privacy policy up on the site. When in the checkout form it links to the generic Shopify privacy policy. No hints to the fact that uploaded images are processed by third parties, as far as I can tell. That should be corrected for sure.

pietz2• 1 day ago

Are they comparing 3.6 Flash to 5.6 Luna and losing? That's ruff.

polski-g• 1 day ago

Why wouldn't they? Luna isn't a Flash model. OpenAI hasn't released a flash-equivalent model since gpt-oss-120b.

pietz2• 1 day ago

Did you ask me a question and then answered it yourself in the very next sentence?

Anyway, given that both Gemini and OpenAI have 3 sizes of models, one would think Google compares their medium size to OpenAIs.

primaprashant2• 1 day ago

Pricing per million input/output tokens:

2.5 Flash: $0.3 / $2.5

3.0 Flash: $0.5 / $3

3.5 Flash: $1.5 / $9

3.6 Flash: $1.5 / $7.5

---

2.5 Flash-Lite: $0.1 / $0.4

3.1 Flash-Lite: $0.25 / $1.5

3.5 Flash-Lite: $0.3 / $2.5

jjice• 1 day ago

Am I off, or does Google have the pricing that varies the most between model generation releases?

m_w_2• 1 day ago

It seems that they're trying to push up-market, or at least they were.

Given the extremely competitive releases of GLM 5.2 and DeepSeek V4 (both pro and flash), I don't think there'll be appetite for it.

LaurensBER• 1 day ago

Pricing often reflects what the vendors (expects) the customer is willing to pay. It seems that Google is still trying to find their niche in the market.

mchusma3• 1 day ago

3.6 Flash would be a great model at 3.0 flash pricing. At this pricing, its thoroughly trounced by about 10 models on cost/performance including Grok 4.5. 3.5 Flash-ite would be a great model at 2.5 flash-lite pricing, as is, its trounced by many models including Deepseek v4 Flash.

As is, they are thoroughly outclassed for most usecases. I will say the one area where i do see Gemini punching above its weight class is in tasks that are effectively "Google this for me" / knowledge stuff. So it does have a role, and I do use it. So while I think Google is still in a strong position overall, they are really stuck as a tier 2 AI player right now with text models. They are tier 1 in bio, images, and video.

Melatonic• 1 day ago

Could also be that they are pricing it at levels where they actually make money. Without seeing the behind the scenes compute cost on all of these its hard to really judge.

That being said with any open model we of course do know the total cost (or estimate)

SwellJoe2• 1 day ago

3.5 Flash was always too expensive for a "flash" model. They marketed it as "near frontier" level, but there are several order-of-magnitude cheaper open models that compete with it.

XCSme8• 1 day ago

In my tests, 3.6 Flash is NOT more token efficient, so it actually ends up costing more than 3.5 Flash, even with the output price reduction.

EDIT: It less less verbose in final output though, but it reasons more.

I assume the optimization comes when you have long-running tasks with many tool calls, and by reasoning more, it reduces the number of tool calls needed.

zuzululu4• 1 day ago

2.5 flash was the only reason we were paying four digits a month to Google....

i guess we'll use 3.0 flash but thats going to get replaced too right ?

these flash lite models aren't very reliable or consistent

jeffybefffy519• 24 hours ago

I wonder if this is a plateau towards the real pricing of AI, if you layer in gemini-2.0-flash at $0.10 / $0.70 then its a 15x price increase to 3.5/3.6 flash. But it hasnt gone up again which is interesting.

sidcool• 19 hours ago

Why are output tokens costlier?

holistio• 1 day ago

They are comparing against their own previous models instead of competitors. Not a great sign.

ChrisArchitect• 1 day ago

Some more discussion:

Gemini 3.6 Flash https://news.ycombinator.com/item?id=48993130

ComputerGuru• 1 day ago

So 3.6 Flash is a somewhat of an admission that Google miscalculated by charging 3-5x for 3.5 Flash what it did for 3.0 Flash (3x input and output costs plus large token inefficiency changes) despite only modest improvements?

3.5 Flash Lite is only a hair cheaper than 3.0 Flash, but I think 3.0 Flash is a massively more capable model?

tiahura• 1 day ago

3.5 Pro must really suck.

WarmWash10• 1 day ago

The mention of an "ambitious" gemini 4 pre-train signals to me that 3.5 pro is probably a lost cause.

That being said, it seems that Gemini is still the best image analysis model, so hopefully 3.6 flash builds on this even more.

canergl• 1 day ago

2 red flags

1- no comparison with gemini 3.1 pro

2- no comparison with any other model

gs17• 1 day ago

The model card has comparisons with both 3.1 Pro and other models:

https://storage.googleapis.com/deepmind-media/Model-Cards/Ge...

youssefarizk• 1 day ago

3.5-lite is the real showpiece here; agentic models of this size are a huge value-add for 90% of knowledge work agent tasks

sreekanth850• 1 day ago

Google is walking backwards, with such a pile of cash in pocket, i feel they are doomed.

thevinter• 1 day ago

I struggle to see any value in this when DeepSeek is still a thing.

anthonypasq3• 1 day ago

multimodal + latency

kzrdude6• 1 day ago

It's smarter than DeepSeek v4 Pro (preview) on several benchmarks, like HLE.

WhitneyLand• 1 day ago

The silence is deafening.

Google watches over the last few months a flat out assault on the Pareto curve from American and Chinese companies. Release after release pushing the boundaries of frontier intelligence and price/performance.

And the response from arguably the biggest AI research labs in the world by headcount is Flash 3.6.

What do you do when you are given essentially unlimited resources and still find yourself falling behind?

ece2• 1 day ago

Start your own openrouter.

cubefox2• 1 day ago

I wonder whether this is more the fault of Hassabis or Pichai. They are clearly both less capable than Altman or Amodei.

kthinckley• 1 day ago

Google desperately needs to make some leadership changes within their Gemini team now that they've been surpassed by 3-5 open weight models and risk loosing frontier status all together in the near future.

alephnerd• 1 day ago

Open weight models aren't likely to be open weight in the long-term. China has started considering export controlling and limiting access to model weights [0].

[0] https://www.ft.com/content/6049a031-9e9b-464c-97bb-414da04d5...

ErneX• 1 day ago

That contradicts this:

https://www.wsj.com/tech/ai/chinas-xi-touts-open-source-ai-a...

So who even knows.

logicchains3• 1 day ago

They don't need to be open weight in the long term; once there's an open-weight Fable-level model with 1M context it'll be pretty much good enough for all coding tasks, no need for new models.

llmslave2• 1 day ago

I keep saying this and people dont believe me, but I have b2b saas systems with actual agents running around the clock, and the performance/stability of the flash model is higher than most other models.

Meaning, its predictable with tool calls, wont spin off a million tools/do weird behavior, its reasonable. Even sonnet in a real world decision making scenario is not reliable, or will reason so long its incredibly expensive.

The benchmarks arent catching all the value, and most people have never actually ran an ai agent in a real context that matters

sureMan6• 1 day ago

Who's most people? What are you talking about? Most people here use agents every day and I wouldn't trust flash or pro to touch any important project of mine because they're both terrible compared to the competition, waste of time every time I give them a chance

llmslave2• 1 day ago

I mean like an ai agent doing some sort of HR work, not a coding agent. Very few businesses are trusting an autonomous agent.

onlyrealcuzzo2• 1 day ago

Gemini 3.5 flash is already a pretty good model. But, unfortunately, the primary way you can interact with it for coding is through Antigravity - which is actively developer hostile.

It doesn't matter how good the model is if you're (mostly) forced to use it in Antigravity - which turns any model into crap.

Wake me up when Antigravity doesn't suck.

postalcoder• 1 day ago

I wonder how big the Pro model is that Google is using behind the scenes to train these smaller ones.

Going on baseless speculation, the lack of accompanying pro models with these flash releases either means: 1) the model is too big to be economical, 2) google doesn't have the compute to serve the big model, 3) their big model has too many alignment issues to serve to the public.

edit: looks like benchmarks are up on https://artificialanalysis.ai/models/gemini-3-6-flash. It's solidly middle-of-pack. However, if you want to be most fair to flash, look at the intelligence vs time per task and intelligence vs outputspeed benchmarks. This is a very fast model.

edit 2: I use antigravity from time to time and in my experience, 3.5 flash is an underrated model, so long as you know what it's good for. It's very good at frontend (much better than gpt 5.5) and it's fast, so it's a great tool for iteration. I expect 3.6 to be no different.

petercooper• 1 day ago

I wonder if the broad use of AI overviews on Google search results is having an impact. Maybe the numbers make it more profitable to use their compute on several billion searches a day rather than selling API access.

WarmWash10• 1 day ago

I think it's a safe bet that Google seems more interested in making a model that improves Google rather than making a model that improves workers.

Fast, light weight, ok intelligence. Perfect for serving 20B+ prompts per day mostly surrounding banal human things.

OAI and Anthropic's cloud spend can cover the revenue gap, as Google is already capturing a large chunk of those guy's revenue.

neutronicus2• 1 day ago

The AI mode on Google search is pretty impressive. Helped me figure out what a bunch of stuff I was seeing out the window was while traveling.

SadErn• 1 day ago

Microsoft also seems to be working in this space. They recently released this:

https://huggingface.co/microsoft/bitnet-embedding-0.6b

It’s a small multilingual embedding model designed for things like search, RAG, and semantic similarity. It supports a fairly large context window and is designed to run efficiently on a CPU in a GPU starved world.

The interesting part is that it builds on BitNet, using ternary weights of -1, 0, and 1 instead of the usual floating-point weights. That should make indexing and searching large amounts of text much cheaper without giving up too much accuracy.

paxys• 1 day ago

AI overview is just a summarization of the top 2-3 results. Of course at Google scale that will still need a ton of compute, but the requirement for generating an overview is many orders of magnitude lower than asking the same question in Gemini.

bitshiftfaced• 1 day ago

Not to mention internal use cases, such as prediction-related tasks like serving ads.

jauntywundrkind• 1 day ago

Or perhaps 4) it's outcompeted severely by other models & releasing it would only tarnish their name

spyckie22• 1 day ago

I wonder if they waited for the new TPU generation to train a larger base model.

Tenoke• 1 day ago

It's also very possible that they know their big model underperforms chatgpt 5.6 and fable by too much, so they are focusing on what they can get wins in like speed instead.

verelo2• 1 day ago

This is the feeling i get too. Cant produce quality, but can produce something that is super fast...so take the wins where they are.

mediaman4• 1 day ago

There was some recent reporting that a July release of the Pro model got pushed back for exactly that reason. Its performance was not good compared to the OpenAI/Anthropic big models. They are having a lot of problems with posttrain.

maxloh2• 1 day ago

I personally doubt that.

It would be a shame if they cannot beat Kimi K3 or Qwen3.8 Max, both of which are claimed to be Fable-like. If that is true, it will be [or would be] the first time a major American lab falls behind a Chinese competitor.

martinald2• 1 day ago

Yes agreed - I wrote this up a while back https://martinalderson.com/posts/whats-going-on-with-gemini/

My view then was they are optimising the models for inference ability on their own hardware AND use cases, which is often speed and time to first token.

They've somehow seemed to end up with terrible compute shortages, which again is surprising given how good Google is at infra deployments AND have their own hardware. From rumors out there they are turning down enterprise deals for Gemini because they don't have the compute.

The problem is they're falling further and further behind on frontier class on coding especially, and since I wrote that article it's got even worse with open weights models undercutting them on price AND intelligence.

janalsncm3• 1 day ago

That and/or the business case isn’t as clear when serving enormous models? You’re constantly stuck in a red queen’s race where your profitability window is increasingly measured in weeks because the Chinese are right behind you.

For small models (which are probably distilled from their big ones) you can serve them economically all the time and not hemorrhage money.

ignoramous• 1 day ago

focusing on what they can get wins in like speed instead

Speed as a differentiator has always been Google's thing. They (used to?) show the microseconds it took to query & rank web-scale search results. Chrome, notoriously, focused on speed at the expense of resource use. The very many efforts to efficiently speed up Android & its runtime since its inception, and so on...

their big model underperforms chatgpt 5.6

Possible but TFA claims:

  We have started our most ambitious pre-training run yet, for Gemini 4 ...

godwinson__4-8• 1 day ago

Didn't they already acknowledge this?

Paywalled article, but the headline is basically all you need: https://www.bloomberg.com/news/articles/2026-07-16/google-ge...

Miraste• 1 day ago

That's the only explanation that makes sense. If it was frontier but cost or compute were limiting factors, they'd release it at an obscene price for the bragging rights. Google doesn't care that much about alignment, and I don't think it's likely to be significantly different than 3.5 anyway. The only reason it would need to be soft-canceled is if it's terrible, and has to end up in a ditch like Llama 4 to avoid shareholder panic.

tpm3• 1 day ago

"3.5 pro is testing with partners! will hopefully land soon."

https://x.com/OfficialLoganK/status/2079596415509303596

awongh6• 1 day ago

It seems like there are some credible rumors that Google is actually winning in terms of actually building models that work and don't lose money- between how they're able to price them, the TPU advantage and their capex advantage (being able to raise debt + just having a lot of cash - well I said not lose money... more like not go bankrupt).

From the outside they look like they're behind in terms of frontier models, but I think they might be the best positioned to not go out of business when the bubble pops.

Also look at the fact that they've been able to deploy AI-assisted search at google scale. It must be another order of magnitude larger (at least) than the model deployments for OpenAI and Anthropic.

Of course unless you're inside Google it's impossible to know for sure.

redox99• 1 day ago

They basically don't exist in the currently most profitable LLM market (coding).

Yes, subs like codex are heavily subsidized. But API billing has massive margins and that's what enterprises pay.

dTal2• 1 day ago

In terms of open models, Gemma 4 beats the pants off everything else to the point that paying for APIs becomes hard to justify. Qwen has the meme-share for coding, but it feels much less well rounded. I have no doubt that Google have both the infrastructure and the expertise to curb stomp everyone else, should they resolve in earnest to do so.

Lest we forget, "Attention is All You Need" came from Google.

deltaqueue• 1 day ago

That "TPU advantage" might be slowing Google down (though likely not as much as their internal bureaucracy).

Porting CUDA-based research, debugging, and overall experimentation speed is likely slower.

The GPU is still king for training.

onlyrealcuzzo2• 1 day ago

It's rumored that Gemini 3.5 flash has a >50% margin, and I'd imagine 3.6 flash is even higher.

I do not think OpenAI or Anthropic are actively chasing margins - though, Anthropic is supposed to be profitable on some form of non-GAAP accounting...

I suspect Google isn't really interested in seeing how far it can get dragged into a race of selling dollars for $0.25, and is more interested to see if it can stay in the race selling $0.50 for a dollar - when everyone else is losing or barely breaking even.

anthonypasq3• 1 day ago

Logan Kilpatrick said on an interview not too long ago that flash 3 and 3.5 are the same pre-train. all gains on top of 3 flash are post-training

mchusma3• 1 day ago

Maybe, but they said they have “started” the Gemini 4 pretrain. So not having done any significant pretrain in a year or so seems odd to me.

re-thc• 1 day ago

the lack of accompanying pro models with these flash releases either means:

Rumors say 4) it didn't perform well, especially in coding so has been delayed

ocamoss• 1 day ago

Maybe it's like Meta not releasing the big version of Llama 4 a year or two ago

SXX2• 1 day ago

I choose fourth option.

4) googles big model just performs worse than K3 and GLM so they choose not to embarass themself.

Like I love Gemini and use it a lot to one-shot whole MR with huge contexts, but its just much worse when its come to tool use and agentic coding.

tonyhart74• 20 hours ago

or they just don't want compete in coding space ???

they have search,youtube,android,office suite like gmail,maps,spreadsheet etc

coding is the least of their problem/priority

rjh292• 1 day ago

I think it's 2. I frequently get told there's no capacity for Pro and the query is answered by Flash with extended thinking. And tbh it's hard to tell the difference between the two, especially if you're not coding with it.

hn8726• 24 hours ago

It's hard to tell the difference because they nerfed Pro to oblivion, it used to be much, much better model (even for non-coding/chat)

zwaps4• 1 day ago

More likely they don't manage to advance benchmarks on the SOTA level anymore. In other words: They can't beat 5.6 nor Fable

joshu• 1 day ago

2.5 flash was absurdly capable on a cost basis

atif089• 22 hours ago

Friend works for Google vendor who generates data for training. His team alone is 200 people (in US).

He says there are many similar vendors and teams with thousands of people in India and other countries.

steve-atx-7600• 20 hours ago

Artificial analysis always seemed sketchy as hell. If you read some of there methodology you’ll see a lot of <=3 repetitions on a particular pass for a given model. So low for calling a frontier model over the public internet ????

retinaros• 19 hours ago

Flash versions were often ultra competitive and their best in the range along with openai mini models. Always been gemini most useable and best model with nano b. Frontier is much more competitive. Anthropic haiku is like 2025 flash…

smcleod• 17 hours ago

I suspect the entire 3.x family is fundamentally problematic and we'll need to see an architecture change before they're half decent like back in the 2.5 days again.

reacharavindh• 15 hours ago

I wish someone would convince Google to may be leave the Google search be without AI responses and use all their resources for a Gemini subscription/API..

bjackman• 13 hours ago

This is a very fast model.

I was already impressed by how fast 3.5 Flash was. But I've never compared it to other models in its class for coding.

Why? Coz models in that class are not very useful to me. Time saved waiting for responses usually just turns into time wasted replying to low quality responses.

Google need to release a Pro model ASAP. I am skeptical of the "maybe they don't have the compute to run it" thing. Anthropic were (probably) in that situation with Mythos and they announced it anyway - that's the obvious play for investor relations as well as hype for your product.

Kinrany• 10 hours ago

3) their big model has too many alignment issues

This implies that normally models are aligned and there is merely a number of issues to fix.

Zigurd2• 8 hours ago

I've been using 3.5 flash in Android Studio on a Dart/Flutter project, some of it pretty complex and using brand new API's and native code for agentic tool calling in the app.

It used to be that you could find the edges of the training set pretty easily. No longer.

_alternator_• 7 hours ago

What I think is really going on is an attempt to segment the market in favor of Google's strengths. It's a bet that models are "good enough" for many use cases even before they reach human-level intelligence, and Google is trying to capture workflows where quantity beats quality.

They are likely deliberately avoiding the SoTA race for a few reasons:

1. Their best models are marginally better than current SoTA releases. 2. They'd like to let Ant/OAI make mistakes with safeguards / let them get the regulatory heat. The unknown unknowns are huge with SoTA models (eg OAI accidentally hacking huggingface) and they are protecting their reputation. 3. They want to encourage companies to become cost conscious because they can likely win on price in the long run. Getting market share in "quantity beats quality" workflows forces companies to establish processes to choose the "cheapest acceptable model", which is a good environment for Google.

parsimo2010• 1 day ago

Feels like they released this to ride the wave of press of GPT-5.6, Kimi K3, and Qwen 3.8. Doesn't feel like Google has much substance with this post except a bump in version and tweaked their pricing.

primaprashant2• 1 day ago

A couple tidbits:

Beyond today’s releases, Gemini 3.5 Pro is currently testing with partners and we plan to make it broadly available as soon as it’s ready.
We have started our most ambitious pre-training run yet, for Gemini 4, and are excited by the progress.

cubefox2• 1 day ago

I guess they meant to release Gemini 3.5 Pro shortly after 3.5 Flash, but then Mythos/Fable and later GPT-5.6 came out with higher performance than 3.5 Pro, so the managers decided not to release it.

jdiff• 10 hours ago

That reasoning didn't stop them from releasing this batch of models, admittedly that may be less face to lose away from the flagship position.

kingstnap• 1 day ago

While them fixing token bloat on 3.5 Flash is good work. That paragraph was the real highlight.

Hopefully 3.5 Pro is soon, and that Gemini 4 can be here end of year and finally have an updated knowledge cutoff.

prox• 10 hours ago

What is a “knowledge cutoff” ?

—Ah, got it, it knows more about recent times.

zb3• 1 day ago

we have taken an intentional approach to deploying 3.5 Flash Cyber. The model will be exclusively available to governments and trusted partners

Screw your government! US and Israeli governments should get the least access, but of course we all know they'll be the (only) ones to get full unfiltered access.

m4tthumphrey2• 1 day ago

I'm going to get downvoted/flagged but I feel like we need a new type of "Show HN/Tell HN" etc for "New AI Model Available".

Front page is tedious these days.

tremarley• 1 day ago

If you refresh the Home page, once a day. The top post will likely be 'New AI Model Available '

Alifatisk3• 1 day ago

I'm going to get downvoted/flagged but [...]

Why do you care that much? Just say it. You're letting an imaginary score determine if you should express your suggestion for improvement. That's a bit wild.

m4tthumphrey2• 1 day ago

No I'm not, I expressed it...

I prefaced it with that to simply say I knew it'd be an unpopular opinion.

ece2• 1 day ago

Just switched to AI Plus from Pro, seems like I won't be missing much.

spyckie22• 1 day ago

Google seems to have anorexia when it comes to model intelligence. They have an internal hard constraint on price per token it seems, and they are trying to squeeze out intelligence with limited compute.

I wonder if there is something with their TPU cycles that makes them want to postpone training a new model. My guess is that they have been on the same base model for 6 months and they may have waited for the next gen TPUs to train Gemini 4, which greatly limits how much intelligence they can increase and forces them to do cost efficiency increases.

WarmWash10• 1 day ago

Google Cloud is probably Google Deepminds biggest competitor. Big company kinda bullshit.

platinumrad• 1 day ago

How so?

logicchains3• 1 day ago

I'd guess they did model-hardware codesign but the design ended up limiting the scaling capability of the model (i.e. they overoptimized too soon).

JacobAsmuth6• 1 day ago

Could it be that they have to serve their models to billions of users?

ur-whale5• 1 day ago

Could it be that they have to serve their models to billions of users?

And how is that different from their competitors exactly?

Zigurd2• 8 hours ago

It's interesting to see a competitive landscape shift out from under the early adopters. Anthropic appears to have anticipated this, but everyone else is going to get caught out by the major platform providers' multitude of distribution channels and use cases. Meta might also be an exceptional case if they can find a way to goose up ad effectiveness.

zmmmmm2• 21 hours ago

It's probably a mix of things but I do think they are viewing "edge AI" as their strategic play: on-device, small efficient models (Android / iOS) and instant AI summaries in google search etc. So all of their focus is on delivering strong performance in a compute constrained environment.

I do think it's still also simultaneously true that they have an actual problem with competing with current frontier progress. It's just that has gone from an existential threat to something they are willing to defer addressing because they see the long game for them sitting at the smaller end.

AussieWog93• 1 day ago

A lot of disappointment here in the comments, but models like these aren't meant to compete with the likes of Fable or GPT 5.6.

I use 3.1 Flash Lite regularly to classify listings on eCommerce websites. It's great for this task - fast, cheap and accurate.

In fact, it was the single best model we tried in terms of the speed vs accuracy vs price tradeoffs - including the Chinese models.

Of course, 3.5 Flash was more accurate but the 5x cost increase couldn't be justified.

3.5 Flash Lite sounds like it could be a strict upgrade for our use case, without a significant increase in costs or drop in speed.

It's not GPT-6 but it's not trying to be. It's a completely different tool and great at what it does.

fur-tea-laser• 1 day ago

not a google fanboy by any stretch... though i've been thrilled with the flash line of models... i exclusively use it on high, and have found it to be a great fit for increasing productivity 10-fold while maintaining quality... sure it can't just go off and one-shot a bunch of work, but at the complexity level i tend to work at, neither can the frontier in a robust way that i can be confident in... sure i have to be in the loop more, but that helps keep me grounded and course-correct earlier before wasting tokens... and when you sufficiently spec out a coding/software problem, and i mean really document all of the critical nuance, it will successfully satisfy the constraints... the quality is rarely acceptable on first-pass, but it forces me to stay connected to the architecture more than i would be if using a frontier model... i've found this to be a happy middle-ground of productivity and awareness...

swe_dima2• 1 day ago

It's scary relying on Google's models.

I have a very price sensitive workload that used to run on flash 2.5 lite - it's deprecated now.

The replacement 3.1 flash lite is a lot more expensive, but now also has a sunset date.

3.5 flash lite is even more expensive.

So the price is rising and you have no choice but to keep paying more and more.

rayboy1995• 1 day ago

I moved directly from 2.5 flash lite to deepseek v4 flash, its already cheaper and if your prompt caching is good you can save so much more money.

binary132• 1 day ago

could you explain how to optimize prompt caching or point to a doc about it?

hagen8• 1 day ago

Just switch the model, its not that much effort tbh. And u can also get a cheaper model than 2.5 lite for the same intelligence

tacoooooooo2• 1 day ago

its not always that simple. dropping in a new model is trivial, but highly specific workflows may rely on specific _invisible_ aspects of a model. when that model gets deprecated, the workflow needs to be rebuilt/re-tuned to work with a different model.

google's inability or unwillingness to provide stable timelines for model deprecation makes it risky to build complex workflows using their models

ActivePattern• 1 day ago

You would be surprised how much of a difference the model makes for certain niche tasks.

For my use case, `gemini-3.1-flash-lite` is ~20% higher accuracy than the next best model of comparable cost (considering both proprietary and open-weight alternatives)

deaux2• 1 day ago

"Intelligence" being what, math? Coding? Unfortunately there's a billion use cases for LLMs whose performance is not at all captured by the popular benchmarks they're all trying to maxx.

zuzululu4• 1 day ago

same I just switched to OpenAI after using flash 2.5 lite for almost everything at our company. We spent thousands just to build this workflow now Google says screw off

ElFitz• 1 day ago

That’s part of why, since Firebase, I’ve tried to never depend on Google products for business, especially not GCP.

Features stay in Beta for ages, whatever that actually means, and released ones get deprecated things fast.

Where some of the competitions treats deprecating entire services as "let’s not put it on your frontpage, put deprecation notices all over the doc, and politely ask new users not to start new project with them".

superkuh• 1 day ago

I felt the same way about openai's text-davinci-002 and code-davinci-002 (gpt-3.5). They were amazing completion models and openai basically dumped them with no equal cost or equal performance replacement. Instead all their models are opaque with no ability to work in completion mode where one actually controls the text input to the model.

These days no company even has completion models where one controls the text input fully. Worthless.

raducu• 1 day ago

So the price is rising and you have no choice but to keep paying more and more.

I presume you can't use deepseek?

bradfa• 1 day ago

There are plenty of 3rd party providers hosting deepseek models, if you don't want to use the 1st party API. 3rd party providers are generally slightly more expensive, but still quite cheap compared to other models of similar vintage and size.

karolist• 1 day ago

sadly it's not multimodal

Cyclone_• 1 day ago

They know that there's big enterprises that will have a strong preference to work with another big enterprise instead of relying on a younger company. At least that's why I think they believe they can do this sort of thing and get away with it.

viccis• 1 day ago

So the price is rising and you have no choice but to keep paying more and more.

You can also just write code like you did a year or two ago.

57016524002• 1 day ago

same here. our production workloads was on Gemini for 2 years. seeing Google unilaterally dropping perfectly fine models and charing you 50x more for worse results is not good.

we are switching to Deepseek.

h2aichat• 1 day ago

Opencode Go is just the same. Each month I will I can do less. Dont ask me why?

pdntspa• 1 day ago

I'm running price-sensitive data extraction workloads on flash 2.5 and its still the king when it comes to accuracy + cost, all the gemini 3 variants perform a bit worse and cost a lot more. Low-key freaking out, ngl

thinkingtoilet• 1 day ago

All models are increasing in price. Everything up to now has been subsidized by investors, private and public.

greatgib• 1 day ago

And somehow, the most annoying is not even the price hike, but it is that is you expect to build a product on any of theirs models, they spend their time being deprecated and you have like to be on the lookup to start from scratch selecting a model and fitting it every year or so... Impossible to have any stability...

simonw4• 1 day ago

Pelicans for 3.6 Flash and 3.5 Flash-Lite (Cyber isn't available to me through the API yet.)

https://tools.simonwillison.net/markdown-svg-renderer#url=ht...

rednb• 1 day ago

I am growing tired of these pelicans posts every time a new model is published. Feels to me like low effort personal brand promotion. Just sharing my 2 cents.

tomrod• 1 day ago

Its a nice benchmark. Like hearing the ice cream truck on a summer day.

cayley_graph• 1 day ago

Yeah. It's something I can do myself in a couple seconds if I want, also on more varied SVG scenes. If this is going to be a benchmark people turn to I'd like to see more effort put into it than just a one-sentence prompt.

IshKebab• 1 day ago

Yeah and it's surely in the training data by now. Long past time to stop.

GaggiX• 1 day ago

I like seeing the pelicans, it's a tradition.

FuckButtons• 1 day ago

My 2 cents: you don’t have to look at the pelican if you don’t want to.

x187463• 1 day ago

At this point, it's kind of a hackernews thing. Simon posts them as a single comment in the relevant thread. It's okay for this place to have a little bit of a sense of community, and you can just ignore the comment.

bayganyo• 1 day ago

I feel the same way. It was fun at first but has gotten tiresome. Does anyone actually use these models to generate SVGs?

busymichael• 1 day ago

I think you're underweighting the Pelican test.

Not only does it give you a super easy-to-grok understanding of the model quality just by looking at the image, but when you compare tokens and costs (both input and output), you really get a good, simple COST x QUALITY evaluation across models.

Simon explains it well: https://simonwillison.net/2026/Jul/16/kimi-k3/#what-can-we-l...

Simon, you should put up a summary table page that you update after every release.

bubble_niter3• 1 day ago

A new model arrives. The pelican, with uncanny commercial instinct, is never far behind.

Sponsored blogs and paid newsletters are after all, notoriously poor at subsisting on silence :)

simonw4• 1 day ago

You and a few other people, but enough people still appreciate the bit that I'm going to keep doing it.

They're easy enough to skip - click the little "-" icon and you'll collapse the entire sub-thread.

theowaway213456• 1 day ago

But how else am I supposed to know when we've reached AGI, until I see an absolutely flawless pelican?

All of the pelicans so far have had really weird flaws / quirks so I am always a little interested to see how well these models perform at this task, since I've seen all the past pelicans and have some anchoring.

Seeing a truly flawless pelican would tell me that the model has true visual reasoning capabilities as well as good taste.

magicalhippo• 1 day ago

Perhaps freshen it up and extend the test by feeding the model the rendered output so it can iterate once. Assuming a multi-modal model.

squidbeak2• 1 day ago

Disagree. They're a nice tradition, but besides that, they're a useful way of eyeballing improvements. I realise labs are likely to be training for Pelicans - but if they're all training for them, the differences in the results are as indicative as they were before labs trained for them.

The 3.6 Flash pelican is just about the best I've seen.

BeetleB• 1 day ago

I love them. Keep 'em coming.

netdur• 1 day ago

Do something instead of complain

EstanislaoStan• 1 day ago

Vibe code an extension that autocollapses any post mentioning pelicans and by simonw?

peder• 1 day ago

All the models do this well. It's a test that tell us nothing at this point.

andybak• 1 day ago

I'm happy for Simon to post what he wants, when he wants. He's earned it.

nullgeo• 1 day ago

Hard disagree. I love a little bit of whimsy (which I feel the world is lacking more and more everyday) from Simon everytime a new model is announced.

reinitctxoffset2• 1 day ago

I'll split the difference. When it's a blog post there's usually an interesting observation or two, but if it's totally automated? Maybe just do the ones with a post.

justinhj• 1 day ago

I find Simon's work informative and entertaining; the last thing he can be accused of is low effort. The Pelicans are just a bit of fun icing on top.

rjh292• 1 day ago

It's just how he is. Prior to LLMs he was cramming a datasette link into every thread. Downvote and move on.

zuzululu4• 1 day ago

Your comment reads very pedantic with a hint of jealousy. The pelican and xbox controllers are great ways to see how well it can follow direction dealing with svg a difficult format for LLMs to use and testing their spatial vision awareness.

jwrallie• 23 hours ago

Here is a different opinion. I’m always looking forward to see the pelican whenever a new model is released. It’s plain simple to understand, memorable, subtle enough in terms of details, and my favorite part is that you have been doing them consistently long enough for it to be useful for comparing almost everything with anything.

theshrike79• 2 hours ago

I want to start a new benchmark: "hacker news commenter saying they are tired of pelicans on bicycles" :D

algoth1• 1 day ago

Flash-lite did the John Cena Pelican

vinaigrette• 1 day ago

I generated a very stylish Pelican using the webapp. Hard to put a judgement on it relative to yours https://share.gemini.google/XSfmve2mEGDV

irthomasthomas• 1 day ago

A piece of the frame is missing between pedals and back wheel. The frame of the bike passes through the bird. It also puts a cap on the bird's head, and a fish in it's mouth.

The fish and the cap where always added when I asked an llm to improve it's first attempt.

This continues the trend in LLM progress of better=more stuff

Edit: I wonder if this is a function of the reasoning training, where more tokens/ stuff is rewarded.

purple-leafy2• 1 day ago

You should be banned for your constant spamming of this. Its ridiculous. Every single AI post! Constant personal promotion.

ishurand42• 1 day ago

3.1 Flash Lite has a better pelican that 3.6?

jbrooks84• 12 hours ago

We need a mod to delete these self promo messages

spprashant• 8 hours ago

I only see a hat under 3.5 Flash-Lite. Is it a model issue or a rendering issue?

andromaton• 5 hours ago

impressed at the look (not the geometry) of 3.6 lol at flash-lite could not find your posts about them

lambda• 1 day ago

3.6 Flash scores exactly the same as 3.5 Flash on the Artificial Analysis index. Better on some tasks, worse on others. Mostly within what I'd consider the noise window. Looks pretty much indistinguishable from 3.5 Flash, at least on these benchmarks: https://artificialanalysis.ai/models/gemini-3-6-flash

theplumber• 1 day ago

I think it’s safe to say Google seems a bit out of the top AI competition now. The “cyber” stuff also starts to become laughable with open models providing the full power without the crap Anthropic, Google, OpenAI are trying to frontload on you(I.e you are not allowed to develop/review a login system, pay a special cyber operation team to do it for you). They really deserve to become irrelevant in the future of AI.

metahost• 1 day ago

So about the same “intelligence” as Muse Spark 1.1 but 2x faster and about 2x as expensive.

stonewhite• 1 day ago

Google somehow managed to snatch defeat from the jaws of success with their AI products.

They literally forced me and my company out of Antigravity by phasing out AI Ultra subscription without any proper product follow-up. Antigravity IDE cannot even have poweruser subscriptions now from Google Workspace an Gemini Enterprise Agent Platform cannot be attached to Antigravity IDE.

Gemini Enterprise Agent Platform has an incredibly abysmal setup process, and if I want to limit spending per-user I have to create projects per user. The fact that you cannot activate Anthropic models on it if the billing still has free credits is almost a joke.

I was a big proponent of Google and Gemini, but they left us reeling with their abrupt product decisions. Forced us to buy $200 subscriptions directly from Anthropic/OpenAI.

ngrilly2• 1 day ago

Also a big proponent of Google and Gemini, but their stubbornness in artificially splitting their consumer and enterprise products is extremely annoying. It's pretty weird that I have access to more powerful tools when using my personal Google account compared to my corporate Google Workspace account.

vel0city2• 1 day ago

As someone who's been using Workspace as a personal email account for over a decade this has been such a struggle forever. Just lots of odd limitations to feature sets all over the place.

When they swapped Google Assistant for Gemini as the default voice provider in Android Auto it was so annoying. My wife's non-work space account can get Gemini to do the normal things like play music and what not, but my Workspace one can't do much of anything at all. I can talk about nearly any random topic with it, but getting it to change the playlist, nah, can't help you there.

It's no surprise to me to see them fumble actually supporting a lot of the consumer features of Gemini into Workspace.

deaux2• 1 day ago

It's literally the meme of the MS org chart pointing guns at each other.

The GCP team wants their slice, the other team wants some otjer slice, and so on. Everyone wants some crap for their promotion package.

It's no wonder Meta has shit the bed even worse.

It's also why Google still releases actually decent, useful models despite the product being such a hilarious mess. A lot of the time Gemini models have actually been better as production LLMs as part of LLM-based production applications than OpenAI and Anthropic models when it comes to the complete cost:quality:latency:adherence picture. And they still are. We have products in prod that use Gemini because they're better than any other model at the specific task. But we wouldn't dare use it for anything coding related, or even just as productivity tool to rely on, because as a consumer product it's a joke.

scrollop• 1 day ago

Apparently you cannot turn off using your data as training data with gemini. This is in line with Google's general privacy policies and it's seeming need to create a stasi file on every human.

reaperducer• 1 day ago

They literally forced me and my company out of Antigravity by phasing out AI Ultra subscription without any proper product follow-up

That sounds awful.

For those of us who don't follow the AI hype cycle, what does that have to do with the topic of this thread: Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber?

browningstreet• 1 day ago

From the posted link: "3.6 Flash: Our workhorse model that delivers better coding, knowledge work, and multimodal performance."

What is Google's recently released AI coding product?

da_chicken• 1 day ago

That's kinda like asking why someone might complain about Chrome's memory usage in a thread about a new version of V8.

logicchains3• 1 day ago

You don't see what Google phasing out an AI product subscription has to do with a Google AI product release?

zhengyi13• 1 day ago

"Your new models are great, but uh... How exactly am I supposed to use them?"

That is, given a lot of users' contexts, the way they can and the way they want to use these models are increasingly disjoint.

ralusek• 1 day ago

It's funny that they triggered the infamous "Code Red" moment in OpenAI when the 3-3.1 models came out. I switched to using them for a lot of single-shot LLM calls because they were fast and cheap. Their only area that they were lacking in was agentic/tool calling.

Needless to say 3.5 was a disappointment. Curious to see 3.6.

xnx3• 1 day ago

Google somehow managed to snatch defeat from the jaws of success

This is still very early days. Who is "on top" has flipped back and forth many times already. The next frontier model release (from whomever) will change things again.

onion2k• 1 day ago

I don't think that's true if the reason a company has left the vendor for given model by making it hard to buy. Enterprise IT is enough of a pain in the butt that people will forego the new shiny to avoid the old painful unless it's genuinely better. As you say though, the best frontier model flips regularly, so companies won't go through the hassle of deploying a model if it's proved horrible to do in the past. They'll just skip that model because their current one is fine.

Eridrus3• 1 day ago

I don't think it's that early tbh, agentic coding has ~90% adoption in the US.

Claude Code has largely won individual developer mindshare and has been on top ever since it came out. The benchmarks change, but almost nobody opts to use anything other than Claude IME when I ask them. Enterprise is more competitive since they care about costs and other things, but developers leaning towards Claude puts a thumb on the scales there.

The product doesn't have much lock in, so it is possible to dislodge Claude, and Anthropic could (and some may argue is likely to) just shoot themselves in the foot again and again and again, but Google has never been particularly good at enterprise sales, and they have never actually been at the frontier of intelligence.

I think Google's incentives have mostly about building models for their products, which makes them focus more on the cheap end, and while they need that, it feels like the Innovator's Dilemma is biting them here.

I own a lot of Google stock from working there in the past and have been quite happy about their trajectory up until the last 6 months, but I am getting pretty antsy about their AI story these days.

creshal• 1 day ago

Model quality is only one aspect, the bigger problem is making it work in a fully integrated enterprise platform, and Google has always been lacking when it came to the latter.

At this rate, if Google has a flagship model, you're better off plugging it into a competitor's tooling than hope Google figures out how to use it.

piyh• 1 day ago

Early days or not, Google fucking deleted my IDE and wiped my settings. It took them days to roll out a fix, by which point I had migrated off Antigravity.

noodlescb• 1 day ago

I guess? Maybe I'm alone here but I don't feel like Fable is particularly more useful than Sonnet most of the time. I feel like the LLMs are good enough for the majority of uses and the hyper expensive premium ones are way into diminishing returns. At this point with Kimi being as good as it is, if they jack up the price any more I'll just go open source.

GodelNumbering2• 1 day ago

I was a big proponent of Google and Gemini, but they left us reeling with their abrupt product decisions.

Likewise. This seems like a common feel. I have at least spent $4000 and likely a lot more on Gemini API because I really wanted them to win. I gave up.

furkansahin• 1 day ago

I am going to ask a very direct question and only because I am curious.

Why do you care? Why would you spend your own money to a multi trillion dollar company so that they win their own "war" against another multi trillion dollar company?

Please don't get me wrong, I know the question can seem a bit negative, I am really just curious.

repeekad2• 1 day ago

Aren’t the subscriptions extremely subsidized and burning cash for Anthropic and OpenAI? A reasonable explanation is they’re simply abstaining from the war of attrition, especially given cheaper comparable models are breaking the illusion that the “frontier of intelligence” has any kind of per token margin.

lerchmo2• 1 day ago

Makes sense, subsidizing tokens doesn’t seem like a great strategy for a public company.

Certhas2• 1 day ago

This is hotly debated and completely unclear. Let's say Anthropics Opus models cost the same to serve as GLM 5.2. GLM 5.2 is 4.4$/MTok while Opus is 5.6 times more expensive. Assume that GLM 5.2 is served at essentially zero margin. Then Anthropic has >80% margin on API pricing. So even if an average person with a subscription pays only 20% of the API price of their usage, Anthropic makes money on subscriptions.

And the real numbers could be better for Anthropic. It's feasible Opus models are actually cheaper to serve than GLM 5.2 because Anthropic have optimized the hell out of inference.

bdcravens2• 1 day ago

Possibly, but aren't the tech giants positioned to win a war of attrition? Then again, they're more likely to sit on that cash and wait for the opportune time to buy a frontier lab.

LUmBULtERA2• 1 day ago

The subsidizing thing is repeated over and over without proof. Personally, I doubt they're actually subsidized.

lanthissa2• 1 day ago

they're subsidized if you max them out, i'd imagine most users are paying $20 for maybe $2-5 of tokens.

anthropic probably has more customers that use more of their sub, but for open ai where a lot of their subs are consumers through chatgpt.com, they have a lot of free money to work with there

zmmmmm2• 21 hours ago

I think the subscriptions pay them back in spades because the same dev who maxes out their subscription on their personal account transfers that exact behaviour over to their enterprise work and - guess what - it's all billed per token there. This is a large reason why corporates are reeling from the cost right now, I think.

deanc• 1 day ago

It's absolute insanity. They have all the resources to have been able to lead from the front with this new technology. They have the products and users to integrate this technology into people's already existing lives. But they keep fumbling.

They're not even benchmarking against other models now, just against themselves - which tells you everything you need to know.

msabalau• 1 day ago

What does "leading from the front" get them?

There is absolutely no loyalty when it comes to coding. Nothing could be more common than people threating to jump ship whenever another frontier or open source model comes.

Google is clearly able to keep growing their free and consumer and small business use cases. Unlike corporate coding, we actually have evidence that solo and small businesses can actually see productivity gains.

Anthropic and OpenAI need to stay dancing like mad, because it's their source revenue which underpins their investments.

Why does Google need to shove something at the top at the same desperate cadence? Other than "recursive self improvement leads to AGI" it seems perfectly fine if they push out something dramatically better every year and half.

LogicFailsMe• 1 day ago

They simply don't have the leadership to lead. And it starts from the top.

asdfman123• 1 day ago

Google promotes people based on shipping features, not making good software. So this breakage means someone is shipping features.

You should be happy for them.

SkitterKherpi• 1 day ago

They'll probably be back later. It's very possible they are "saving up" for a much bigger run.

bdcravens2• 1 day ago

More likely an acquisition.

notatoad• 1 day ago

i've got to wonder how much of this is intentional, and how much of this is just google being their usual terrible selves at anything consumer-product related.

Google's biggest and most important customer for all this AI stuff is google. Do they actually want other customers, or is having other people use their AI just an annoyance at this point, where we use up compute that they'd rather use internally...

wolvesechoes5• 13 hours ago

google being their usual terrible selves at anything consumer-product related

And yet millions of people around the world are using their stuff, very often using only their stuff.

nxdmum• 1 day ago

Google Gemini agy is not allowing you to use your token via your own harness. My own harness is far more efficient than agy. They can take a simple stance - if you exceed your token limit they block you - with the 5 hr limit they are already doing this . so there should be no reason to block you from using your own harness - if you are more efficient - you gain - if you are less you lose .

They are not allowing me to hit their endpoints which agy hits - it's frustrating . i tried to hack it with gemini itself. what i love about gemini is it's so encouraging and ready to help you - even against the agy client : ) .

Even though im so frustrated with this - i still love Gemini for some reason ! Most encouraging model in the world!

dainiusse• 1 day ago

+1 same way vscode killed its usage base and gave it to its fork - cursor

epolanski• 1 day ago

Google somehow managed to snatch defeat from the jaws of success with their AI products.

HN lives in a bubble.

I have German/Italian/Polish clients virtually all use Gemini and NotebookLM. Talking insurance, banking, consulting, legal.

The real world doesn't look at pointless benchmarks on writing react tailwind crap, they are already google suite users, get the tools, test them and adopt them, end of story.

It's going to be like with angular, never mentioned on the net, widely used in the real world.

LUmBULtERA2• 1 day ago

I have friends and family who use Gemini, but entirely because their Pixel phones came with a year of it for free. No other reason, and they will most likely never pay for it.

wolvesechoes5• 13 hours ago

Yup, it is often mind-boggling how people here are detached from reality. Millions of people around the world rely on Google services, Gemini in all probability is the most used model consumer-wise, Google receives unimaginable stream of data on even smallest habits common people have etc.

But hey, Gemini makes shittier job than Fable at producing my shitty "app" no one cares about. It is so over for Google!

ikiris• 1 day ago

Snatching defeat from the jaws of victory is the specialty of product managers.

ur-whale5• 1 day ago

they left us reeling with their abrupt product decisions

This is what happens when you put a McKinsey consultant in the role of CEO of an organization where product managers run the asylum instead of engineers.

giancarlostoro• 1 day ago

I trust Google to:

* Abruptly ban me and all users from a Google Workspace for no reason whatsoever.

* Abruptly shut down a service.

I'll never give Google my direct money.

throwuxiytayq2• 1 day ago

Life gets simpler and better as soon as you stop giving money to Google.

It’s so silly that individual people still use their shit. Corporations, I understand - they always choose the most mediocre stacks and tools by default. But why people choose to bring the mediocrity of Google into their lives is beyond me.

redml• 23 hours ago

I don't know where you've been for the last decade but all google ever does is retire products with no recourse.

ansuman441• 1 day ago

Specific to task these can be huge plus point.

dismalaf• 1 day ago

With all the naysayers on Gemini models I'm curious how many people actually use Gemini regularly?

For me, Gemini models are the most usable. Claude Opus and Mistral always try to turn queries into one-shot enormous commits, which just burns tokens, time and annoys me for something which is still wrong more often than not.

Gemini seems far better at listening to instructions and giving me what I actually want, on top of using far fewer tokens and wasting my time. Fable is the only model that's come close to Gemini Pro for me.

And as this is about Flash, it's exciting, I find Flash can usually get the right answer pretty quickly and without too much nonsense.

dudeinhawaii2• 21 hours ago

I use all of the major providers daily and I tend to go to Gemini for "fast lookups" where a good enough answer is probably OK. I use ChatGPT and Claude for anything where it matters and generally when I invoke all three -- Gemini is the most surface level with responses, and also sycophantic.

It gets worse from there. Gemini is terrible at agentic coding, primarily because Agy is terrible. I noticed Google updated Agy with this release, so perhaps that's finally going in a good direction. I'll have to test it. Thus far, my experience in countless experiments has been Gemini models being 2x faster yet with less depth in their solutions and a lot more going off track.

I very rarely have to stop Codex or Claude Code sessions because they're doing something random and unexpected (or not asked for). I genuinely think Gemini models are brilliant but virtually useless in agentic scenarios in my experience of the last few years (2.5, 3, 3.1, 3.5).

I should also note that Gemini web UI annoying resets to its lowest intelligence which feels scummy and Google is not transparent about what "extended thinking" really is. Past posts have pointed to "extended" being medium. Every other provider gives you the raw value (medium/high/etc).

So honestly, I feel Google would rather I don't use their models. They just want to get a little bit of mindshare and stay in the conversation. I had the Ultra plan and cancelled it once it was apparent they were not improving the agentic experience nor trying to compete.

tobias2014• 20 hours ago

I agree, I see how Gemini itself with a usable harness can be excellent. But in agy with forced eager compaction (~125k with 3.1-pro, ~200k with flash) a kind of laziness and forgetting shows through that leads to an endless sequence of stopgap instructions, even with rigorous GEMINI.md and isolated task delegation and a good task tracking system. Agy is basically useless for more complex problems as far as I am concerned, at least when used somewhat autonomously as one could expect from claude. For strictly mechanical one-shot tasks it might be fine. I've spent way too much time working around these limitations instead of just continuing to use claude. Hoping that things would have improved with flash 3.6 I feel that it's actually worse in following instructions, and always acts even when just asked a question. If just agy offered a better experience and got rid of the terrible forced automatic eager compaction.

PS: That opus-4.6 via agy works so much better points in another direction though!

xnx3• 1 day ago

Proof-of-life release while they figure out how to have a competitive frontier model release. My hunch is they pushed too far in the "omni" model direction, that they made something so ungainly, it wasn't as good for normal tasks.

Gecko4072• 1 day ago

I read this as a soft let down to not expect too much from 3.5 Pro.

We have started our most ambitious pre-training run yet, for Gemini 4, and are excited by the progress.

zuzululu4• 1 day ago

Google seems to be falling way behind the pack. antigravity cli is pure trash. gpt 3.5 pro is now behind and isn't released yet. GPT 6 and Fable 6 releasing next month. What the hell is going on over there ?

zarzavat2• 1 day ago

What the hell is going on over there

Google was late to coding agents and as-per-usual fucked it up with their crazy project management culture.

Usually Google gets away with it due to inertia, however this time they are paying a heavy price because they missed out on the training data that Anthropic and OpenAI have gathered with claude and codex.

accountrequired2• 1 day ago

whatever, dude. give gemma5

revolvingthrow2• 1 day ago

Tons of guardrails, lazy model, super confusing plans, expensive 3.5/3.6 flash and lite and 3.5 pro MiA?

Rough patch for google ai

gabriel-uribe• 1 day ago

Haven't been excited for a Gemini release since December. Wild to see.

XCSme8• 1 day ago

I was expecting 3.6 Pro. It's been so long since the last Pro model...

thebigspacefuck3• 1 day ago

They are working on coming up with a better code name. You know, something like ”Fable” or ”Sol”, gotta have one these days. Personally I think they should go with “Mafia”. How cool would that sound? 3.6 Mafia.

dpacmittal• 1 day ago

3.6 Gangsta Pro

baalimago• 1 day ago

Not good enough for high-end, not cheap enough to be for low-end. Next!

u1hcw9nx• 1 day ago

Google has not changed. Following two facts are like tautologies by now.

1. Their AI efforts are very fundamental research oriented. They are really good at it.

2. Their productization sucks. The end products gets little attention compared to competition. It can be canceled at any time. You should never build anything around Google only APIs, AI or not.

mythz2• 1 day ago

Always happy to see new Gemini releases as IMO Antigravity Pro 16.67/mo plan (Annual) is still the best plan available and have been pretty happy with Antigravity IDE.

If it wasn't for Gemini/Antigravity I'd have to go with a Max Claude plan, as it stands now I can get by with just a Claude Pro plan to get Opus when I need it, whilst using Antigravity as my day-to-day workhorse.

Unfortunately Gemini Flash became too expensive to use as a general purpose model (i.e. for AI features in Apps), luckily there are plenty of cheaper Chinese models to fill that gap now.

ValentineC• 1 day ago

Why do you think it's the best plan available?

JacobAsmuth6• 1 day ago

(Rate limits * capability of the model) / cost

Andrex3• 1 day ago

What's the current outlook on Antigravity IDE vs. 2.0? How long will they begrudgingly keep it going before kicking everyone to 2.0/3.0?

(I actually use a mix of both for some offline projects, nothing serious.)

mythz2• 12 hours ago

Antigravity is now split into 2 Apps:

'Antigravity' which is an agent-first editor layout where you don't see your code and just prompt it.

'Antigravity IDE' which uses the Windsurf/VS Code editor, which is still what I primarily use in my day-to-day.

I hope they never retire the IDE, I don't think I can get used to prompting an AI Agent without being able to see my code to help workout what needs to be done.

lenerdenator2• 1 day ago

We're almost five years into the whole GenAI thing and we're still relying on these guys to spoonfeed us incremental updates.

It's time for them to start focusing on open-weight models and efficiency. Otherwise there's just a layer of marketing hype and "will it do this?" that has to be cut through for evaluation of each and every release cycle.

Models are getting easier and easier to create. The money, if there's any here, is in the harness the user interfaces with, and the data centers running them.

raffael_de• 1 day ago

is it just me or is this one-upping each other every few days getting ridiculous secreting a whiff of desperation?

JacobAsmuth6• 1 day ago

Just you. This is typical market competition in a fast moving field.

vlad_recomply• 1 day ago

Models are expensive and low performance. On top of that they make you jump through hoops to even use these models without being throttled even for the weaker models. The only reason we are using them is credits. As soon as credits run out we are switching immediately.

QuesnayJr• 1 day ago

I remember back when Gemini looked like it was the best model that this comment section was full of confident predictions that Google had "won" and that no one would ever catch up with them again. The most embarassing part is that I kinda believed them.

summerlight• 1 day ago

Looks like 3.6 Flash is the first model with their newest pretraining run (cutoff date is 2026/03), long after 2.5 series.

XCSme8• 1 day ago

tl;dr: 3.6 flash is a bit smarter than 3.5 flash, but also a bit more expensive.

My results [0] put Gemini 3.6 Flash at the top.

3.6 Flash high has same $1.5 input price as 3.5 Flash, but output is cheaper from $9.0 to $7.5.

Google said 3.6 Flash is more token efficient, but in my tests it's actually LESS token efficient[1] than 3.5 Flash, so despite the output price reduction, it still costs more.

[0] https://aibenchy.com/compare/google-gemini-3-6-flash-medium/...

[1] https://aibenchy.com/compare/google-gemini-3-6-flash-high/go...

sagex• 1 day ago

Don't know why are they even pursuing Gemini. Just download the Kimi, call it Kimini and serve it on your GPU. Maybe then train next architecture based on this!

thebigspacefuck3• 1 day ago

IMO Gemini has the best free tier models/app for everyday use. Muse-Spark is perhaps just slightly better, but has none of the connectivity to my GApps (for things like “create a recipe in my Google Docs from this image”).

Plus they are probably running these things on every Google search so saving tokens is a huge win for them.

copperx2• 1 day ago

Free? Did I misread the pricing details?

thebigspacefuck3• 22 hours ago

Free plan, the default tier without requiring a subscription. If you use through the Gemini App or gemini.google without paying anything, the model used is 3.6 Flash.

Rankings for text are here https://arena.ai/leaderboard/text

For comparison of Free Tiers: - Gemini serves 3.6-flash (rank 12) - ChatGPT serves 5.5-Instant (rank 23) - Claude serves Sonnet 5 (rank 27) - Meta AI serves muse-spark-1.1 (rank 5)

While Meta AI serves the better ranked model, it doesn't end up working that well for other things. For example, if I ask "help me buy a new raincoat", it ends up suggesting a Cambodian website, whereas Google is well integrated with Google shopping. It doesn't have the same integration with GApps outside of Gmail/Calendar. A few other email connectors are available.

Claude has one of the best interfaces with connectors, skills, and plugins galore, but the model and limits are restrictive on the free tier.

Gemini, as far as I know, I've never hit a rate limit on Flash.

I believe Gemini is going to gain market share through the free tier funnel while serving models as cost-effectively as possible. People are going to use Gemini because they use GApps and Google.

ChatGPT and Anthropic are going to be competing for the API/Business users, but for everyone else they are going have to become Google before Google becomes them.

TheAtomic• 1 day ago

I have liked using their consumer products but they don't make it easy, that's for sure.

game_the0ry• 1 day ago

At this point, I think google should consider becoming a hyper scaler for anthropic and open ai, and I predict that that is exactly what they do. The model is no longer the most valuable part of the stack.

dakolli• 1 day ago

I use 3.5 flash 10x more than any other model, despite have access to all of them. If I'm going to play a slot machine, I'd rather get the pain over with quickly.

maxdo• 1 day ago

quite a good model, the speed/price/quality ration is a new golden intersection for me, not sure if its as good as grok 4.5 but quite fast/capable model.

JeremyHerrman• 1 day ago

Gemini 2.5 Flash-Lite has been my go to for cheap document processing at scale (especially with 50% off batch mode), but they are really boiling the frog with pricing increases with each version:

gemini-2.5-flash-lite: $0.10 input / $0.40 output

gemini-3.1-flash-lite: $0.25 input / $1.50 output

gemini-3.5-flash-lite: $0.30 input / $2.50 output (a 6.25x increase over 2.5!)

Now watch them deprecate Gemini 2.5 Flash-Lite in the coming months...

tjwebbnorfolk• 1 day ago

gemma4 is the same price as 2.5-flash-lite, and performs better.

JacobAsmuth6• 1 day ago

How has your experience been with Gemma 4?

luciana1u• 1 day ago

the real product is the naming confusion we made along the way. Gemini 3.6 Flash, 3.5 Flash-Lite, 3.5 Flash Cyber — at this point even the model cards need a model to explain them

Havoc3• 1 day ago

Flash Lite: 0.3/m and 2.5/m

Deepseek Pro: 0.435/m 0.87/m

That's wildly ambitious pricing by Google. You can maybe get away with spicy pricing at the SOTA edge but at the lower tiers everything is a lot more price sensitive.

JacobAsmuth6• 1 day ago

You need to compare cost per task buddy boy. Cost per token doesn't tell you much when you don't know how many tokens a model will use to accomplish a task

Havoc3• 1 day ago

boy

Seriously?

spstoyanov• 1 day ago

Glad to see the price is going down but it's still too high for a "fast" model

57016524002• 1 day ago

if only DeepSeeek supported vision, would never use Gemini.

vinhnx• 1 day ago

For anyone wanting a faster overview: I ran the Gemini 3.6 Flash and 3.5 series release notes through NotebookLM and generated a short video summary. Link: https://www.youtube.com/watch?v=SUFBhvQ2tY4

speak_plainly• 1 day ago

It feels like AI is going to be the end of Google. The post-Schmidt company culture cannot produce consistent, consumer-friendly products that any sane person would want to use consistently.

fuomag9• 1 day ago

no actual cyber model release, useless

lukewarm707• 1 day ago

"The model will be exclusively available to governments and trusted partners via CodeMender soon as part of a limited-access pilot program"

we are stealing plutocracy from the jaws of emancipation.

i don't want to live in a world where abundance is guarded and shared among politicians and cronies, whilst the rest are left to rot.

MILP• 1 day ago

I'm a big fan of the Flash-Lite models. They're exceedingly fast and deliver great outputs for high volume use cases where you need to process requests at scale. Can't wait to try the newer version.

imagetic• 1 day ago

If only I could use Pi.

kzrdude6• 1 day ago

Well, you can use the google models from Pi. Go to aistudio.google.com and set up an API key. There is a free quota, it's relatively large for the Flash Lite models.

...and last time I looked the limits were more generous for Gemma 4 there, but they have been tightened a bit. That's how it goes, always changing.

goldenarm• 1 day ago

LLM reception is truly extreme, even worse than AAA game releases.

Ever frontier lab lived it at least once : missing the frontier by a few months triggers extremly negative reactions, then you take back the lead for 2 weeks, and the hype cycle repeats.

Andrex3• 1 day ago

I spin a mental roulette on whether the reception on a new release will be "OMG best model by far, no one will be able to catch up for months!" or "OMG this is already outdated, RIP company X, they might as well just give up now, there's no coming back from this."

It's quite a fun game. I click into the comments and see if the roulette wheel was right.

ianberdin• 1 day ago

Pelican svg and a near-perfect 3D MacBook at max effort for $0.16, about a fifth of Fable's price.

Fable 5 still wins on detail with no visible errors, but it's close. And this isn't a memorized pelican;

https://playcode.io/blog/macbook-svg-benchmark#gemini-3-6-fl...

WarmWash10• 1 day ago

I don't know if it's a rendering error since it looks like your site renders the SVG instead of hosting a static image of it, but the 3.6 macbook looks like an abstract art piece lol, both ff and chrome desktop

Arshad-Talpur• 1 day ago

never tried gemini for coding, but this news seems to be compelling, i would definitely give it a try

sega_sai• 1 day ago

I have just tried to switch to 3.6 instead of 3.5 in antigravity and it seems to constantly spit "critical instruction: STOP CALLING TOOLS NOW. YOU MUST WAIT FOR WAKEUP. ". I think I will switch back to 3.5

parasti• 1 day ago

Kind of excited about this. 3.5 Flash on Antigravity has surprised me recently on a hobby project. When given opportunity to plan, it can deliver on tasks that would take me a while on my own and generates responses at blazing speeds - compared to what I'm used to at work with Opus 4.8 (granted I don't use Opus 4.8 on my hobby projects so just anecdotal). While with Gemini CLI I would just watch it run in circles and run out of 5h allowance before anything useful is produced (or even approached).

lilytweed• 1 day ago

Really, what's up with Gemini still not supporting connectors/MCPs/plugins/whatever-they're-called-this-month on web? It makes it a non-starter for any kind of serious use.

nicce• 13 hours ago

Why would you use them on the web? Serious use happens elsewhere.

vrosas• 1 day ago

I have no skin in this game and this comment will be gray in a few minutes BUT a friendly reminder that these types of threads are astroturfed heavily by competitor labs and any info should be taken with a massive grain of salt.

lwansbrough2• 1 day ago

Plugged 3.5 Flash Lite into an existing agent harness that was previously using 3.1 Flash Lite and this shit just does not work. It's not following instructions and is not producing the correct tool calls.

kzrdude6• 23 hours ago

Maybe this is relevant? Just in case

For autonomous subagents with tool calls, code execution, or multi-step reasoning: set thinking_level to "medium" or "high" to prevent premature tool termination.

I just happened to see that in the docs: https://ai.google.dev/gemini-api/docs/latest-model

lwansbrough2• 22 hours ago

Thanks, I'll give that a try.

Alifatisk3• 1 day ago

In other good news "the model has been trained to minimize refusals for beneficial uses.".

Otherwise, this news feels like a tiny incremental improvement on Gemini Flash series to make it more efficient with token usage, subagent and cost. Nothing big.

Regarding their benchmark scores on CyberGym, I wonder why they didn't compare their 3.5 Flash Cyber model with Fable 5. I mean they included Mythos and GPT-Cyber, so why not Fable 5 too?

They also mentioned Gemini 3.5 Pro is in testing and its about to become available very soon. Another thing maybe worth discussing is the announcement of pre-training Gemini 4. Sadly, not much technical details to discuss on. Many comments in here seem to mostly be about how Google is behind the others, but honestly, is it really worth the investment to be #1 in Artifical Analysis every week?

ur-whale5• 1 day ago

Why exactly are they announcing these completely milquetoast models ?

I'd be low-keying the release if anything, given how lame they are compared to their competition.

What am I missing?

arjie2• 1 day ago

Their naming scheme is confusing. Branding has never been Google's strong suit and their marketing copy is pretty bottom-of-the-barrel[0]. Anthropic has a pretty clear set of models but Gemini decided to rebrand their Flash as Flash Lite (and presumably the future will see a Flash Lite Mini, a Flash Lite Mini Nano and a Flash Lite Mini Nano 3B) which confuses the pricing to high hell.

This plus the Vertex, AI Studio, Gemini, Antigravity. It's honestly too confusing to use. I need to use Gemini just to decide on which platform and which model to consider.

0: Famous Kurian Tweet: "We're announcing Duet AI for Google Workspace will now be Gemini for Google Workspace. Consumers and organizations of all sizes can access Gemini across the Workspace apps they know and love. We're introducing a new offering called Gemini Business, which lets organizations use generative AI in Workspace at a lower price point than Gemini Enterprise, which replaces Duet AI for Workspace Enterprise."

mrandish• 1 day ago

I often use Gemini free web chat because it's generally quite good at web search-related questions (apparently it has direct token-level access to the Google Search index) but I noticed in the last two weeks output quality of 3.5 Flash seriously degraded. Maybe they were switching over systems.

culi2• 1 day ago

Google's Knowledge Graph is a massive advantage no other competitor has. I don't think they've fully utilized its full potential but I don't know if any other company could've built something like Scholar Labs

Andrex3• 1 day ago

Knowledge Graph + automated transcriptions of almost every YouTube video = giant untapped moat of data

1saadcodes• 1 day ago

Nice to see that it's cheaper than 3.5

mchusma3• 1 day ago

Wow, Laguna S 2.1 (released today) just destroys Flash-Lite underly and completely. What a weak and embarrasing release from Google.

hmokiguess• 1 day ago

Spent half an hour just now benchmarking it against my current 3.5 Flash pipeline excited only see it regressed slightly (0.1% - 0.2% at most, for feature extraction work)

Seems like this is mostly a cost play by Google, hoping this doesn't bring 3.5 Flash capabilities to an end of life, and that 3.6 catches up or gets better.

waldrews• 1 day ago

3.5 Flash-Lite seems available in US region, as was 3.5 Flash; but 3.6 Flash looks Global only so far when pinging. If Google employees are watching, will this issue go away?

jdthedisciple3• 1 day ago

Bottom line it looks about on equal footing with GLM 5.2 in terms of both overall intelligence and cost per task, while being significantly faster (in fact it is the fastest model on artificial analysis as of rn [0])

[0] https://artificialanalysis.ai/models/gemini-3-6-flash

brap• 1 day ago

From my experience, this thing is crazy fast.

Spawn 10 on the same problem and have them debate to reach a consensus, you’ll get Fable-like results but 100x faster.

zwaps4• 1 day ago

Here's the issue:

GLM 5.2 is better, also cheaper, and almost as fast.

So essentially, a big L for Google. Combine this with them not being able to produce a frontier model this generation... hmm implications

WarmWash10• 1 day ago

3.6 is roughly 50% faster, which isn't totally insignificant for being marginally more expensive.[1]

[1]artificialanalysis.ai

zwaps4• 1 day ago

Sure, but there's no sota alternative from Google. That's it, and its beaten by GLM 5.2 on every measure except somewhat speed.

I find that quite staggering. GLM is open weights

HDBaseT2• 21 hours ago

Counter-point, cost per task is almost the same ($0.47 vs $0.50) between GLM 5.2 and Gemini 3.6 Flash. [0] Not to mention the subscription plans likely produce 10x value compared to GLM 5.2 API, unsure the rate limits on a equal subscription vs subscription, but Google subscriptions offer tons of other value, including 1 year of Free Gemini for Education accounts.

[1] https://artificialanalysis.ai/models/gemini-3-6-flash

zacksiri• 1 day ago

Gemini 3.5 flash-lite is more expensive than Gemini 3.1 flash-lite. Every upgrade is getting more expensive.

michaelbuckbee• 24 hours ago

It's kind of ridiculous how good these are getting. 3.5 Flash lite is pretty comparable to Opus 4.8 (at least for the couple tests I did) while simultaneously being 6x faster and 19x cheaper.

https://fy2zp1ri90.evvl.io/

s3p• 21 hours ago

Not for me personally. While setting up a custom website, 3.5 Flash introduced tons of bugs that Claude had to fix. The website has about 3,000 lines of code spread across multiple files, and Gemini somehow couldn't do frontend changes without breaking things. Sharing my 2c, but I've stayed on GPT 5.5+ and Claude Sonnet/Opus 4.6+. Anything past that from those two have been bug-free, but Google's latest hasn't been.

sajithdilshan• 24 hours ago

Google really needs to get their product strategy together. The discontinued gemini-cli and introduced antigravity-cli which is a downgrade IMO and the sooner they can partner up with AWS and release the gemini models via Bedrock the easier corporate/business which has strict data protection rules can use their models and make them available for internal engineers.

It's a one thing to research and improve the model, but if they ignore the ease of access and multi-availability of their models in different ways they are going to fall behind again.

resonious• 23 hours ago

So it's GLM-5.2 performance for almost twice the price.

That said, the speed looks really good. I think it's competitive with Fireworks's GLM 5.2 Fast, although Fireworks is still cheaper.

prtmnth• 23 hours ago

My hunch is Google is trying to integrate a fast and relatively cheap AI across search and every other surface of their product suite. And for that objective, a model that can move faster while being accurate and cheap enough is more important to them than producing a frontier class heavyweight model.

schainks• 23 hours ago

This. Give me cheap tokens that produce accurate information and the deal is done

verdverm2• 20 hours ago

I've found that having good source material (markdown, dependency source, search results for agents) for the models to draw on significantly improves information accuracy. Definitely worth investing in this side of "harness engineering", don't rely on facts burned into weights

do_anh_tu3• 21 hours ago

Man I love Gemini models but these kind of pricing increase is just insanse. I have a little product and I have to keep increasing the price and reduce the limits because of this non-sense, and they did not even let us use the old models in near future, so I forced to update to the new model with basically no to little improvement because I don't even need that much. Google if you can read this, it okay to release new models and change the price for them, but please please don't kill the old ones like gemini-2.5-flash-lite, because that all I ever need for my little apps with only few thousands of users.

Tuna-Fish• 20 hours ago

Do not base products on models that are not open-weights. Doing it is like building a product on someone else's platform, you are entirely at their mercy, and even when they don't have any reason to hurt you, you are tiny enough that if any policy they want to enact hurts you as a side effect, no-one is going to care.

You don't have to self-host the open-weights model, you just need to be able to source it from multiple providers.

Using the closed vendor models maybe made sense when open-weight models lagged so far behind, but that time is now gone.

do_anh_tu3• 4 hours ago

I just have no choice. My application was tested and benchmarked against a lot of models, but no model can be in the same league as gemini in term of multi-language support and understanding.

satvikpendem6• 20 hours ago

Why don't you just switch to cheaper models? I'm sure DeepSeek is probably enough for you and it's way cheaper. If you want, host your own (or have someone else host) open weight models, I use both embedded and cloud Gemma for some things.

shaism• 10 hours ago

What is your use case?

Have you considered moving to open source / Chinese models?

If gemini-2.5-flash-lite is good enough for your application, you will find even lower cost options with better performance outside of the Google ecosystem.

do_anh_tu3• 4 hours ago

I can't, my main task for it was translation, and no other models in the world can keep up with the multilang support like gemini models, because, no other company in the world has the multilang dataset like google and there might never be one, because google has been collected the internet data for it search engine since the 90s already. And so much has been lost in the history of internet and I doubt there any other company in the world that can go back in time to get those data back.

lostmsu2• 20 hours ago

3.6 Flash has the same performance on artificial analysis benchmarks as 3.5 Flash. So... what... is... the... point?..

lostmsu2• 20 hours ago

Nevermind, I just realized it is point six!

t2ance• 18 hours ago

Waiting for Gemini 3.5 Pro...

csunoser• 18 hours ago

This is the darnest thing - all the testimonies are just jepgs? Did they run out of time to work on the css?

ernestrc• 13 hours ago

Fable or gpt5.6 sol for planning. Gemini 3.6 Flash for executing. Wow, Google is onto something here. I always thought that gemini 3.5-flash was the most underrated model. Let's see how much better is 3.6 flash.

Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber

Related Stories