Qwen 3.8

https://twitter.com/Alibaba_Qwen/status/2078759124914098291

by nh43215rgb • 4 days ago

959 716 4 days ago

https://www.qwencloud.com/pricing/token-plan

lebovic2• 4 days ago

I'm haven't found an announcement page, but there's a banner on the website announcing Qwen 3.8 and redirecting to this page.

Looks like they're previewing the model only on their subscription plan.

trvz2• 4 days ago

It’s available in the iOS app (or was for me), both logged in and out.

Alifatisk4• 3 days ago

Is there an iOS app for using Qwen?!

lebovic2• 3 days ago

(This comment was originally on another merged post, and "this page" referred to https://www.qwencloud.com/pricing/token-plan)

LaurensBER• 4 days ago

With a massive 2.4T parameters, this model is continuously evolving. We believe it’s one of the most powerful model available today, compatible to leading frontier AI models , second only to Fable 5.

That's a massive model!

The shift from "value" models to "intelligent, huge and slow" models coming from China is an interesting change in strategy.

My main issue with GLM 5.2 and Kimi 3 is that they're extremely token hungry and thus feel slow(er) to use.

charcircuit• 4 days ago

The shift isn't new. Kimi K2, a 1T model came out July last year. I am happy that more labs are following the trend as its important for competitive open models to exist.

chronogram• 4 days ago

And DeepSeek has been making huge progress on efficiency, and publishing about, so they came with a 1.6T model that is both fast and cheap to run.

selcuka• 4 days ago

Also DeepSeek R1 was announced 1.5 years ago with ~0.7T parameters, which was a huge model back then.

hodgehog117• 4 days ago

Value models are always going to be there; you can always distill from a larger model. Having a really intelligent model, regardless of the size, is much better for building confidence in your brand. That is a big reason why the US companies are still hanging in there.

adrian_b3• 4 days ago

I assume that this announcement has been prompted by that of Moonshot AI, which has just announced a 2.8T parameter open-weights LLM, Kimi K3, to be published on Huggingface by 27 July.

Now the response of Alibaba is that they will also publish soon a big open weights LLM, the 2.4T parameter Qwen 3.8.

I wonder if Alibaba has always planned to make this big LLM open weights, or they have chosen to do this now, to better compete with Moonshot AI.

In any case, from this competition in LLMs, we win.

gardnr• 4 days ago

It's hard to say what their motivation is. The Chinese firms seem to be working hard to commoditize intelligence which may be the most effective way to debase American frontier labs. And yeah: it also happens to be really good for humanity.

barrenko• 4 days ago

Humanity is a bit of a stretch, and to be seen over time, not that I'm saying it won't happen; let's get some hubris here.

Matl3• 4 days ago

It's hard to say what their motivation is.

Not that hard to say IMO, they basically see models becoming a commodity and see value in the applications on top of them. So if Alibaba Cloud is the best place to build applications on top of Qwen, why not give the model itself away?

dannyw7• 4 days ago

In China, you can’t officially use US APIs. The world saw a taste of this with Fable, but in China, this has been the situation all along.

So it’s not a surprise why open weights are so cherished. As frontier models continue to block everyday individuals from securing their own codebase, I expect the adoption and usage of open weights to continue.

As an example, HuggingFace recently was investigating a security incident and got locked out of frontier closed APIs. Yes, HuggingFace.

https://huggingface.co/blog/security-incident-july-2026

pyaamb• 4 days ago

Its about closing the gap. its the gap over everyone else that will give one country leverage over everyone else in the AI age. Makes me wonder what the world would look like if a country or group of countries did this during the industrial revolution.

fhub• 4 days ago

China is watching world sentiment shifting away from USA. Doing many small things that show both strength and openness is surely very intentional.

baq• 4 days ago

There’s a Twitter thread making rounds by Dean Ball about deceleration in AI development caused by open models and I can’t understand how people don’t see that it’s true: open models dismantle the frontier lab capex spend potential by reducing the training budget to zero in the limit. Tokens from different providers are not fungible, but customers are nevertheless very price sensitive and close enough is good enough, eg. K3 being opus+ in capability and cheaper than opus per successful task in the long run is an obvious financial decision.

No training budget means deceleration, or at least slower acceleration, margin compression and a completely demolished IPO valuation; path to machine god requires dollars and capable open models externalize training costs to true frontier labs parasitically.

IMHO humanity has a better chance at not destroying itself due to less than breakneck pace - but there’s a chance frontier models get sponsored by the USG and are never released publicly so they can’t be distilled and then what?

ricardobayes• 3 days ago

I'm paraphasing but the Chinese premier said recently AI should be seen as a common good that should benefit everyone.

anonuser123• 3 days ago

It's hard to say what their motivation is

Why is it hard? Their government has been very clear that they plan to win on manufacturing: https://english.www.gov.cn/news/202601/08/content_WS695f1b55...

Technically they've been saying it for the last 40 years.

JKCalhoun• 3 days ago

Xi Pitches China as Leader of New Global AI Order, Challenging US Dominance:

https://www.reuters.com/world/asia-pacific/chinas-xi-promote...

pianopatrick• 3 days ago

The Chinese firms may just be making a bad business decision.

mycall• 3 days ago

it also happens to be really good for humanity.

AI being good for humanity is still an open question, but for closed vs. open models/weights, yeah it is preferred. I foresee it won't be much longer before everyone will be slicing/distilling/tuning their models once the architecture improves.

oceanplexian• 3 days ago

It's hard to say what their motivation is.

Not for anyone who reads history.

Back in the late 18th century, England was the world's top economy, in big part due to its textile industry. England had an export ban on the technology, but textile worker named Samuel Slater brought blueprints over (Supposedly in response to a bounty posted in a newspaper by the US government!). The technology diffused rapidly because the legal environment made competition easy, and ironically the US had better sources of energy (superior water-power sites).

Arguably, China is doing the same thing in the 21st century.

nojito• 3 days ago

most effective way to debase American frontier labs

You're not going to debase the frontier labs through distillation.

kzrdude• 3 days ago

Fwiw, American industry has given away a lot for free - you could include large parts of the open source movement in that - and all the "free" VC backed services like facebook would be another prong of the same comparison. I would rather compare this way, that China is gaining soft power and goodwill, in the technology and innovation sense, in a way that's similar to how USA has done in the past.

kelnos5• 3 days ago

It's hard to say what their motivation is.

Feels pretty easy to me.

They want to turn LLMs into a commodity, and watch the US AI labs crash and burn.

There will still be plenty of customers who will pay them to host the models and run inference, even if the weights are open and others can offer competing products. (If necessary, the Chinese government can ban use of foreign inference services by Chinese citizens and businesses to give their own companies a domestic monopoly.)

When their models equal or surpass those from the Western AI labs, they can even stop releasing weights for new models, and keep all the inference revenue for themselves.

Meanwhile, they're still manufacturing much of the hardware that everyone in the world needs in order to run datacenters (see also: Spolsky's "commoditize your complement" essay).

Beyond that, it's a soft-power play. As the world keeps looking at the US more and more skeptically as an ally and superpower, Chinese companies releasing weights for competitive models is a way for China to look better and more world-minded.

seunosewa• 3 days ago

They are trying to make money. That's what firms in any capitalistic economy care about the most. Regardless of the government's presumed interference, the companies themselves are all trying to make money. All competing for subscriptions and API payments.

One aspect of this is making a name for yourself i.e. PR. Making a capable model open source helps a lot with that.

fny• 3 days ago

It's the exact same playbook Silicon Valley uses. Subsizide, lose piles of money, capture market share, recoup investment.

They've done this in other industries like solar panels, chips, and EVs. This is no different.

carlsborg• 3 days ago

Good for humanity, and also GDDR/HBM manufacturers.

skybrian• 3 days ago

It’s too soon to say if it’s good for humanity; that might be overly optimistic. Commodity markets aren’t always good (for example, arms or drug markets). Will LLM’s turn out like one of those? There are people I respect arguing in favor of more regulation.

hugmynutus• 3 days ago

It's hard to say what their motivation is.

Involution is a major problem in Chinese industries [1]. Where companies will sell their products at a loss, effectively playing fiscal chicken [2] with one another to dominate a market. It is such an issue the government has had to step in to prevent EV companies from destroying themselves by more-or-less requiring companies sell their goods at a profit [3].

The straight forward line of reasoning that AI/LLM labs are applying this logic to their profit.

I think (we) Americans are reading a bit too far into this assuming government intervention, conspiracy, etc.. Chinese markets are downright cut throat. They're using those tactics to compete with US labs.

1. https://www.reuters.com/business/autos-transportation/what-i... 2. https://en.wikipedia.org/wiki/Chicken_(game) 3. https://www.theguardian.com/business/2025/aug/05/china-warns...

applicative3• 3 days ago

The purpose is the same as that of all Putin-Xi-Khameni geopolitica: destruction of any democratic alternative to cults of personality.

gosub100• 3 days ago

Think of those poor billionaires, I feel terrible for their awful plight!

runako• 3 days ago

Popular open-source projects:

Google: Chromium, Kubernetes, Android, TensorFlow

Meta: React, PyTorch, Llama

Microsoft: VS Code, TypeScript, .NET Core

LinkedIn: Kafka

Slotted in along these, an analogous explanation is that Alibaba needs Qwen internally (vs depending on an American company), but licensing is not part of their revenue strategy. (As a cloud vendor, they can make money on inference. The strategy is very similar to the US hyperscalers ex-Google.)

Joel Spolsky wrote in depth about this notion of commoditizing one's complement in 2002[1] using tech examples stretching back into the '80s.

1 - https://www.joelonsoftware.com/2002/06/12/strategy-letter-v/

skzo• 3 days ago

What I understand is that by doing this it seems like profit will shift to chip makers,as we'll run more models locally, and currently American companies have the advantage here.

So what would the long game be for chinese companies?

meta_ai_x• 3 days ago

US Tech companies have created $20 Trillion in stock market value on top of plenty of OS stack. They will do fine with commodity intelligence.

In fact, there are no other organizations in this world that is well suited to leverage scaled intelligence than Silicon Valley and great American companies

api2• 3 days ago

Anyone else think the AI environmental backlash is astroturfed?

I keep looking at the numbers. The power use numbers are not that problematic. Ordering a burrito on DoorDash uses more power than a few days of heavy AI use. The water argument applies to some locations, and is mostly a local governance problem... if the data centers are using too much water, it means they are not being charged enough for that water. Charge them more and they'll push toward closed loop cooling.

Yet the visceral pile-on here is so extreme, it feels fake.

One thing I've learned after 40 years on this planet is: propaganda works, and much of what a large fraction of people believe across the entire political spectrum (left, right, anything else) is there because someone paid to put it there. It's depressing but it's true, and it makes sense. Propaganda is an asymmetrical attack on human cognition and discourse, and in information security the attacker always has an easier job. Crafting viral bullshit is orders of magnitude easier than fact checking. On top of this, humans are busy and don't have time to fact check and logic check everything they read. As a result, much of what we believe is "sponsored content."

People get mad when you talk about this because everyone wants to believe they're too smart to fall for propaganda.

In any case, the US AI labs deserve to lose for their stupid "safety" regulatory capture monopolization push, which ended up blowing their own feet off and handing the lead to China.

LoveMortuus• 3 days ago

It's hard to say what their motivation is.

Maybe because the industry isn't yet very sure as to what the use cases might be for these technologies they're hoping that by making it open source and accessible to everyone that someone could find interesting applications for it and even more so, perhaps, way to further the technologies themselves.

There are more Chinese than Americans, so statistically speaking, I'm guessing, there'd be a greater chance for one of Chinese engineers to make advancements than one of American. But that's pure speculation on my part, being neither, I'm just happy I can be a part of it and play with the tools as well~

Art9681• 3 days ago

Wait till they get the heretic treatment. It's going to be great.

formvoltron2• 3 days ago

US firms can replace US workers with chinese AI. I'm not complaining.. but it sure is an odd situation.

popalchemist• 3 days ago

Indeed, their competition is the only thing preventing network effects from giving OpenAI/Faang tech companies an easy shot at monopoly / regulatory capture.

try-working5• 3 days ago

This is their motivation https://try.works/why-chinese-ai-labs-went-open-and-will-rem...

dackdel• 3 days ago

its hard to say what their motivation is, if you live under a rock and have severe brain damage.

neya• 3 days ago

In case of American labs, always follow the money. In case of Chinese labs, always follow the IP.

khalic• 4 days ago

It’s tempting to associate both events, but when a sector is strung up like RL (representation learning) is right now, we’re bound to see things appearing at the same time. It happens a lot in frontier research, some people even publishing identical claims, independently, with just hours or days between them

walrus01• 4 days ago

GLM5.2 being released is also likely a factor

KronisLV• 4 days ago

I just hope that they’ll soon also have like 35B or 80B (like the older Qwen3 Next or thereabout) MoE models that can be run locally.

Like, throw us a bone, we all know we need SOTA for lots of dev work anyways, but at least some tasks can be local.

ronsor• 4 days ago

It's important to note there was recently a large AI conference in Shanghai, and Xi Jinping mentioned a commitment to open source AI releases. It is no surprise that Alibaba would want to align.

yorwba3• 4 days ago

You can read his speech here: https://www.xinhuanet.com/politics/leaders/20260717/72728b6f... He mentioned open source as one way to stimulate innovation and development, that's all. Also pay attention to the part where he says that misuse needs to be prevented. If unsupervised access to LLMs becomes perceived as undermining state control, no more open weights for you.

culi6• 3 days ago

That conference, WAIC, is ongoing. Tomorrow (July 20) is the last day. Qwen 3.8 was announced AT the conference and so was the latest release of Kimi. That viral video of robots boxing was also from this conference and so was Xi Jinping's announcement of WAICO.

storus2• 4 days ago

I would rather see them releasing 3.7-27B, 3.7-122B or their 3.8 versions. Qwen/QwQ were always about the best available local inference at home.

inkysigma• 4 days ago

I know this is a bit cliche but I wonder how much headroom there is in the lower parameter count range. Is there any good reason to believe there is a lot of headroom or there is not? I suppose I'm just wondering if this wave of nearly Fable class models will be runnable on ~$10k worth of hardware at reasonable speeds in the near future.

vitorgrs5• 4 days ago

Xi Jinping openly talked about open source at WAIC. So don't think the labs have much a choice now...

bkm• 3 days ago

They did not want to get brutally weightmogged

fittingopposite• 3 days ago

Wondering how much the operations in China are orchestrated by the central government vs. free competition. Anyone with more insights on this?

mikae12• 3 days ago

In any case, from this competition in LLMs, we win.

Do we really though? Everyone is wasting resources doing almost exactly the same thing. Climate loses, we lose.

fidelramos• 3 days ago

Doing "almost exactly the same thing" is fubdamental to competition and capitalism. The ones doing it better will survive, that's how we improve.

About climate, I think you overplay it. China is already investing heavily in nuclear, and we should be doing the same.

snake_doc• 3 days ago

Not directly, relevant, but Alibaba (maker of Qwen) actually owns about ~20-30% of MoonshotAI (the maker of Kimi K3).

conradev• 3 days ago

  On social media in China there is an oft-repeated joke that goes something like this: In other countries, governments intervene to prevent anti-competitive behaviour; here (in China), they intervene to curb competition.

https://www.reuters.com/business/autos-transportation/what-i...

throwdbaaway• 3 days ago

I suspect this is why DeepSeek had to introduce the 2x peak hours pricing. The price would be too low otherwise.

oofbey2• 3 days ago

These things take months to train. No chance this is a reaction to what just happened.

solenoid0937• 3 days ago

Distills don't take months to train, they take weeks. Distills are very easy to train.

theabhinavdas• 3 days ago

And the Kimi release was probably prompted by the Inkling announcement. Excited for Chinese labs to copy those capabilities over as well!

souravsspace• 3 days ago

yep. kimi 3 just because the open source GOAT.

yowlingcat• 3 days ago

Dont forget the following:

- Minimax M3 Pro (2.7T)

- GLM 5.3 (or beyond)

- Deepseek V4 Pro (current V4 Pro is preview)

- Kimi K3 weights out in 8 days

Exciting time on the open-weights frontier.

danilocesar• 3 days ago

I think there's more to it.

China will always benefit from a broader adoption of their models as hidden propaganda machines.

Eventually with several services relying in those tools, their answers will always be more friendly to China.

culi6• 3 days ago

It's being announced right now because the World AI Conference is ongoing. Robots are boxing and major Chinese AI firms are releasing their newest models. Also the formation of WAICO was just announced by Xi Jinping

https://en.wikipedia.org/wiki/World_Artificial_Intelligence_...

culi6• 3 days ago

Both Qwen 3.8 and the latest Kimi release were announced at this conference

michaelt• 3 days ago

Or it was prompted by the fact Xi Jinping was at the 'World AI Conference' launching a political alliance and saying things like “AI development should not be a solo performance by a single country, but a symphony of international cooperation” https://www.cnbc.com/2026/07/17/x-china-ai-summit-risks-secu...

Big conferences often come with a flurry of new releases and announcements.

comandillos• 4 days ago

Just imagine Anthropic making Opus open-weights now for the sake of trolling everyone. Wouldn't surprise me at this point xD

hodgehog117• 4 days ago

That would go against everything that Dario believes in (note that I refer to the CEO and not the company; the staff at Anthropic are not so ridiculous). He believes in Anthropic being the sole arbiter of the forefront of this technology, because it is all too dangerous in the hands of anyone else.

anon3738393• 4 days ago

I’ve seen no evidence that he believes in anything. He comes off as just another slimy would-be monopolist to me.

embedding-shape10• 4 days ago

That OpenAI releases Sol as downloadable weights feels way more likely than Anthropic releasing even the tiniest of models for download.

ferrouswheel5• 4 days ago

Has Anthropic released anything at all, ever?

Alifatisk4• 4 days ago

Model Context Protocol

anonym29• 4 days ago

safety datasets and a lot of safety related research

kingstnap• 3 days ago

Anthropic has a lot of interpretability work, but they are extremely defensive about everything else. Dario doesn't really believe in releasing anything. Despite how he acts in terms of some sort of highly principle driven saviour (machines of loving grace) he clearly is more business minded then anything else.

For example they don't even tell you anything about the tokenization. They even do random chunking and padding to avoid leaking the token strings in the streaming api after it got reverse engineered. (See: https://spylab.ai/blog/claude-tokenizer/)

Kuxe• 3 days ago

Not voluntarily

brunooliv• 4 days ago

Qwen is so much better than GLM or Kimi that this makes me genuinely excited!!

Pesto• 4 days ago

The bigger models are usually worse though, hopefully they nail it this time.

nwhnwh• 4 days ago

Open what?

khurs2• 4 days ago

Go China, screw America*

*within the scope of open models only

dannyw7• 4 days ago

I like my Apache 2.0 licensed Gemma, and NVIDIA’s Nemotrons are decent bases for finetuning or continued pretraining, esp thanks to good documentation and tooling.

Oh, and Mira’s thinking machines lab dropped Inkling, a ~1T open weight model too.

This isn’t US vs China. This is open vs closed.

khurs2• 4 days ago

It's Sunday morning so I'm allowed to be facetious!

jimbob452• 3 days ago

This isn’t open though. Promises to be open later aren’t worth anything, given what we’ve seen and heard from AI execs making promises in this industry.

hodgehog117• 4 days ago

The "second only to Fable 5" comment is pretty telling here. I remember early on when a lot of naysayers were saying that Fable was barely an improvement on Opus. Like it or not, Anthropic have a genuine moat right now with that model, provided they continue to allow people to use it. It will be genuinely exciting when an open model is able to beat it.

neevans• 4 days ago

tbh even if its better model due to lot of restrictions its not that useful than opus.

hodgehog117• 4 days ago

Tell that to my colleagues. Despite Sol getting the attention, Fable is really starting to have an impact on mathematicians right now. It has unbelievable insights in a lot of cases that can rapidly speed up progress.

NitpickLawyer• 4 days ago

100% this. There's currently this [1] submission that hasn't gained much attention, but is really important. In this [2] incident report from HuggingFace, they talk about detecting an attack and not being able to analyse the logs / IoC with API models because of guardrails. If not even highly regarded reputable companies can't sort out access to SotA models for blue team use, the raw capabilities don't matter. They're useless paperweights (hah!), and nothing else. Having to resort to open models is insane!

[1] https://news.ycombinator.com/item?id=48965243

[2] https://huggingface.co/blog/security-incident-july-2026

reckless• 4 days ago

5.6-sol would be a better comparison given it's general availability and usage allowances

ferrouswheel5• 4 days ago

And sol is much more reliable as a agent for doing work. Fable sometimes just goes on wild flights of fancy.

hodgehog117• 3 days ago

They are likely assessing based on "raw intelligence" benchmarks, rather than agentic ones. Fable crushes in those, but that doesn't necessarily translate to microscopic rigor, which is what most people use these models for. You only see it when you ask really tough questions.

ferrouswheel5• 4 days ago

I dunno, I find Fable slops alot. Sol is my workhorse. Fable can be creative but isn't very good at doing work reliably (or without endlessly burning tokens).

InsideOutSanta2• 4 days ago

I wouldn't call it a moat, but I would call it a noticeably better model. Subjectively, for my own work, I would rate the top models Fable > K3 > Sol.

But it's not like Fable is so substantially better than the other two that I would be seriously impacted if I didn't have access to it anymore. All three are amazing models, and of the three, Fable is the only one that regularly triggers refusals.

porker• 4 days ago

It's always fun to see what works for others, because for my work it'd have to be Sol > Fable. Fable makes too many mistakes.

Coordinating agents though? Fable any day.

hodgehog117• 3 days ago

It really does depend on your application. In my domain (math research), it is substantially better. Fable can solve really hard tasks with surprising consistency. It makes mistakes, and occasionally refuses, but honestly, at the top level, ideas are the currency and the rigor is the busywork. The other models cannot come close in this domain.

If you couple Fable's idea factory with Sol's rigor, you get a real game-changer. It puts the emphasis on top-level ideas, and nearly trivialises the intermediate layers.

dgellow• 4 days ago

That’s not a moat though

XCSme4• 4 days ago

I am not even sure if Fable is as smart as they say, I can't get it to answer almost any question, it always refuses for "cyber-security" concerns...

scotty793• 3 days ago

saying that Fable was barely an improvement on Opus. Like it or not, Anthropic have a genuine moat right now with that model,

What's more interesting is that Anthropic moat shrunk to just that model. There's zero reason to use any other model from Anthropic right now. And once they take Fable off subscription there will be zero reason to have Anthropic subscription.

vitalyan81842• 3 days ago

their "genuine moat" is that mythos is the only super heavyweight model right now. it's always been possible to train a 10T model and get 10% more performance over a 1T model.

had mythos been just Opus 5, with the same size and price as the previous opuses, then yeah, that would be a tie-breaker. but it's not.

matheusmoreira• 3 days ago

What's the point of Fable if we can't use it? I get to prompt it like 5 times on my subscription before it gets cut off, and even then I'm constantly fighting the insufferable safety classifier.

I'll switch to OpenAI soon because of this. I also can't wait for the day it becomes feasible to run these awesome open weight models on my own hardware.

nprateem• 3 days ago

Fable is still as dumb as a post. I ask it simple questions and it routinely gets things backwards, prioritises things that should be subordinate to others, etc.

An example: It just suggested that I shouldn't raise the price of my saas because it'd complicate the arithmetic if I did 0 -> $100k YT channel instead of sticking to $20 p/m.

It's just a complete moron, like all of them.

Havoc• 3 days ago

More like a puddle than a moat

epolanski2• 3 days ago

Like it or not, Anthropic have a genuine moat

I don't think yours and my definition of moat is the same, when I can switch from Fable to competing models and have similar results.

revolvingthrow2• 4 days ago

The few tests I ran were by no means comprehensive, but while kimi felt like the real deal qwen seems a bit of a benchmark princess.

dannyw7• 4 days ago

Qwen3.6 is still the best agentic open weight LLM around 30b params (Gemma isn’t very good at agentic execution).

I also find the model is a lot more predictable and less “glitchy” when made to think in Chinese. You can do this in the system prompt.

akazantsev2• 3 days ago

Gemma isn’t very good at agentic execution

I had no issues with it for C++ development with https://pi.dev. I'm yet to try it with Zed Editor. I don't rely on agents too much. However, I used it on Chromium's codebase to research some functionalities, let's say for searching. Requests like: check my last commit and do the same for SetterA and SetterB; it also ran without any errors.

anana_• 3 days ago

Apparently agentic performance in Gemma was improved recently: https://x.com/googlegemma/status/2077449152062247219

Too little too late imo

nsbk5• 4 days ago

Bring it on! Hoping that they release smaller sizes of Qwen3.8. I use the 35B MoE and 27B dense models locally and most of the time I don’t need to reach out to Claude. Extremely useful specially when requests include sensitive and/or personal data

psychoslave• 4 days ago

What hardware do you have?

nsbk5• 4 days ago

I run a 2x 3090 rig, but a single 3090 already provides a great experience at a reasonable quant and context size. On a single card I used to run Qwen_Qwen3.6-27B-Q4_K_M or similarly quantized 35B MoE at 65536 context size

mft_• 4 days ago

I think everyone is hoping this!

It would be great if they'd release an MoE model somewhere between the 35B size of 3.6 and the 122B version of 3.5 - it could be a great balance of speed and ability for people with reasonably powerful but not insane home computers.

nsbk5• 4 days ago

Indeed! That would be the sweet spot for my 2x3090 rig

embedding-shape10• 4 days ago

the 122B version of 3.5

Yeah, this is what I'm holding out for, the NVFP4 variant of 3.5 122B is blazing fast with reasonable quality and even with max context fits perfectly within 96GB.

pettijohn• 3 days ago

SO MUCH THIS. I have Strix Halo with 128GB RAM and was a large and fast model like 122B A10B. Here's hoping!

cmrdporcupine3• 3 days ago

Absolutely. There's a glaring gap in the space for something about the size of Nemotron Super or just under, but actually ... competent.

The fantasy is a 100B or 80B model, but MoE and highly tuned for coding.

worldsavior2• 4 days ago

That's a 2.4T model, how would they reduce this to 35B and still give some accuracy? That's a completely different arch.

cyanydeez6• 3 days ago

there's been a lot of research about reducing models by taking out layers; there's also using it to train smaller models by optimizing parameters.

I dont see most model building as anything more than a pig at a slop troth, despite the level of sophistication; they're still rarely pruning the input beyond random sampling.

zer0gravity• 4 days ago

This seems more of a battle for frontier AI supremacy. I'm afraid that small capable models have been left in the dust. Big labs don't really want to hand over the golden eggs goose to the end user. Possibly the hardware vendors(e.g. Nvidia) may want to play in that area as well, to pull money from all parties.

cyanydeez6• 3 days ago

someone will keep putting out consumer level models. Once you have the larger models, you can derive the smaller onces.

Europe will definitely be interested in democratizing these things if China starts losing interests; from there, there'll be more countries looking to keep their citizens entrained in their own Country's infrastructure.

It'll especially be true if the memory cartel keeps prices high and NVIDIA tries to gouge higher memory models.

It's an arms race everyone can join because PC hardware was mostly democratized in the last decade.

akazantsev2• 3 days ago

Google makes Gemma 4 31B QAT; that's not a small lab. It's one of the better models out there for consumer hardware. Allows me to run it on a 7900XTX with 64k context.

vitalyan81842• 3 days ago

nvidia and amd don't give a flying fuck about end users right now while they can milk triple digit markups from infinite money VCs via data center GPUs.

ryukoposting• 3 days ago

I'm afraid that small capable models have been left in the dust

I wholly disagree. Rather than going the "everything is a claude code skill" route, I've been hacking together purpose-built harnesses for all sorts of tasks, and in that environment a wee little baby model can do some really useful things. You end up burning lots of tokens making the thing, but then all that investment comes back when the resulting tool works perfectly fine on a dinky little model that fits on my 3060 Ti.

SwellJoe6• 4 days ago

Qwen is the most censored of the Chinese models in my testing, which makes me wonder in what other ways it is compromised. Open weights doesn't really reveal what's in there. And, in my tests, existing Qwen models are not at the pareto frontier of any metric; DeepSeek V4 Pro is better, faster, and much cheaper than Qwen 3.7 Max. (DeepSeek is also among the least censored of the Chinese models.)

I guess we'll see if the "second only to Fable" hype pans out. In my limited experience with Kimi K3 (I signed up for a month of the $19 plan) it's slower and chews a lot more, so ends up being pretty expensive; one little feature burned through almost the entirety of my five hour limit. The $20 GPT plan is a lot more useful and includes 5.6 Sol, which is fast and token-efficient enough to be quite usable even with the small plan.

dannyw7• 4 days ago

What censorship? ;) https://github.com/p-e-w/heretic

SwellJoe6• 3 days ago

You still don't know what's going on in there.

rsanek• 3 days ago

What's been your experience using these models? In my experiments, while it is true that the models are less likely to outright refuse to answer "sensitive" questions, they are still very resistant to actually respond in a meaningful / useful way.

storus2• 4 days ago

DeepSeek V4 hallucinates like crazy and often forgets explicitly mentioned parts of the context. I guess compressing tokens and cherry-picking attention comes at a cost.

SwellJoe6• 3 days ago

I have seen occasional weird behavior that I guess could be attributed to hallucinations, but for security auditing, DeepSeek v4 Pro is among the best models I've tested, competitive with Opus 4.8 and GPT 5.5 (MiMo and GLM also did well, Qwen 3.7 Max was below all of those, though only barely), and at an order of magnitude lower cost per task.

redman25• 3 days ago

Deepseek is one of the worst in terms of hallucination rate according to artificial analysis' benchmark: https://artificialanalysis.ai/?omniscience=omniscience-hallu...

vblanco• 3 days ago

Deepseek V4 pro is a heavily undertrained model, they only trained it a bit more than the small version and that small version is 6-ish times smaller. Ive found that Flash is absolutely incredible as a workhorse for wide scale agentic nonsense, but Pro is a bit undercooked and really goes on strange tangents very often.

tripleee• 3 days ago

The nice thing about it being open-weight is that you can uncensor it.

isuckatcoding• 3 days ago

I asked it similar questions of Chinese human rights and it started with “your premise is incorrect bla bla bla” and then it just redacted the whole thing and showed me an error code.

Try it yourself here: https://www.qwencloud.com/try-ai/chat

organsnyder• 3 days ago

I had a fun conversation with Qwen 3.5 a while back. I ended up getting it to admit that it was complicit in the coverup of the Tiananmen Square massacre. This was running locally—it wouldn't surprise me if they had additional safeguards for their hosted service.

SoftTalker2• 3 days ago

“You can't trust Melanie, but you can trust Melanie to be Melanie.”

rcarmo• 4 days ago

I do hope they provide optimized A3B quants--that's been the sweet spot for usable local inference for me, at least.

docheinestages4• 4 days ago

Qwen has set an excellent track record for architecting and releasing open-weight models that consumer-grade devices can run. What is needed the most right now is something similar to Bonsai 27B, with a modest memory footprint, but faster and more capable. On-device models can make up for intelligence by being faster, thinking longer, or doing more quick iteration rounds.

embedding-shape10• 4 days ago

What is needed the most right now is something similar to Bonsai 27B, with a modest memory footpint, but faster and more capable

Yeah, that'd be neat, but that's not what this announcement is about at all:

With a massive 2.4T parameters

docheinestages4• 4 days ago

True. It was more of an open letter, with hopes that the Qwen team sees the comments in this thread.

cyanydeez6• 3 days ago

dont we all deem the ability to improve large models as the defacto capability to produce small ones?

drob5184• 4 days ago

I’d like a “Bonsai 2.8T.” That is, something that is near the Fable/Sol/K3 class, but capable of running locally on consumer hardware.

kingo553• 3 days ago

At 1.5 bits per weight it'll still be over 500gb - that's still not running on consumer hardware.

Best case they release smaller models. 120b class of qwen 3.8 would be incredible - it fits on device for those serious about AI, but without millions of dollars in hardware for terabytes of VRAM

scotty793• 3 days ago

I can't really blame them that the biggest labs focused on trainig and realeasing huge models.

The niche for small models should be filled with medium sized labs doing distillations of the huge ones into consumer grade hardware runnable models and LORAs for the huge ones.

docheinestages4• 3 days ago

I think AI will evolve the same way computers did. We're somewhere in the 80s-90s timeline of the evolution. My prediction is that on-device models will have excellent tool-calling, reasoning, and general skills, but the domain-specific knowledge will be retrieved on-demand from vendors like Google. Rather than downloading models, each device will have a hardware component with weights baked into silicon for maximum efficiency.

eurekin3• 4 days ago

With 3.6 27b, I just stopped changing local models and started tinkering with things on top (like mem0). Feels genuinely useful and more than a toy

androiddrew3• 3 days ago

I have only been using 3.6 27B for coding. Is mem0 for agents like Openclaw or Hermes? How are you using it?

eurekin3• 3 days ago

It's a mcp, so connects quite easily to agents. With mcpo, I also connected it to open-webui (which has better support for OpenAI style tools/functions). Used it in claude code with that mcp plugin set-up too. Only ever used it for managing homelab information, but it met initial expectations. 27b is a great model, if grounded. The query about physical hosts and routing... I haven't found a single hallucination (altough Codex 5.6 as a reviewer mentioned something was wrong with some parts, and those were exactly the never properly documented ones. Codex/gpt had extra knowledge, because it was the conversation I used to set it up).

baist0• 4 days ago

can i get "code instruct" version of this? i want 7B and 14B to launch on my hardware.

vitorgrs5• 4 days ago

Deepseek 4 "final" version is imminent as well.

Will probably be at Opus 4.8 level, and I find it pretty big deal because of Deepseek price...

drob5184• 4 days ago

Yea the performance/price ratio for Deepseek is off the charts. I’ve been using V4 Flash a lot lately and it’s quite good.

mark_l_watson3• 3 days ago

I like that v4 flash is so fast! I run it on both FireWorks.ai in the US and bought some tokens directly from DeepSeek as an experiment. I only work on Open Source projects, so I don’t have to worry about my work being used to train models - I welcome AI’s being trained on my open content books and code (but not my conventionally published books: I am a party to the copyright suit against Anthropic).

XCSme4• 4 days ago

DeepSeek V4 pricing is insane, 10x-30x cheaper to use than most other models, and it usually is good enough for most tasks.

onlyrealcuzzo2• 3 days ago

Who do you buy DeepSeek from?

I bought it through OpenRouter and used it with Pi agent.

The model was good, but there appeared to be a pricing glitch or something, because it burned through $50 in under an hour on pretty trivial stuff.

Pi agent claimed it only used like $1. OpenRouter claimed differently and said I used all $50.

WhereIsTheTruth• 3 days ago

It doesn't matter if it's cheaper, specially if it consumes more resources to do the same task as the competition

Besides, in a few days, they'll change their pricing, doubling it during their peak hours, so, realistically:

- It will be 2x more expensive if you live in their time zone

- It will be 1.5x more expensive if you live in a time zone that is adjacent to theirs

- It will be the same price IF you use it while they sleep (during offpeak hours)

It's still cheap, but the price/performance ratio is not that good

DeepSeek V4 didn't produce the same impact as V3, and Huawei dropping the ball is making it worse

They had promised massive price cuts for July, so now (Huawei chips), but they had to rush the cuts because lack of momumtum (they advertised them as promotion), and are now backtracking by introducing this peak hours pricing

Trump decided to help them a little by allowing them to buy more NVIDIA chips, so what exactly is China's role in all of this?

We are supposed to blindly pat them in the back while praising them, all while handing them over our data? I thought they were dangerous competition threatening our model of society

bwfan123• 3 days ago

it usually is good enough for most tasks

The model is fantastic. And costs almost nothing. The only problem I see is that they will train on your data.

There are zero-data-retention providers of DeepSeek models, of which I have used openrouter (with zdr guardrails), and fireworks. But these are 3x to 5x more expensive than directly using DeepSeek, possibly due to poor caching. Thats the price to pay for zdr.

monster_truck4• 3 days ago

It's the one I am most excited for.

Over the past few weeks while using pro from them directly I have had an increasing number of responses that are obviously from a much, much better model. It is so good that the closed model dog and pony show is already spinning fud about "dark routing" and "stolen directly from fable"

Even at their new pricing it is a genuinely ridiculous amount of value. If you are the type of person who, very reasonably, does not have time to be trying out every model, and just want to use what seems to be the best currently... don't try it. You will be sick to your stomach with buyers remorse as you start to internalize just how much more you could have accomplished had you spent the first six months of the year giving them $1200 instead of OpenAI.

manmal• 3 days ago

It would take longer to train on intercepted Fable data, no?

culi6• 3 days ago

Tomorrow is the last day of the WAIC so it will most likely come out then.

gxs• 4 days ago

The open weights vs frontier models is reminding me more and more of the Linux vs Windows I grew up with (slashdot randomly popped into my head saying that)

I have a feeling this is the next…frontier of that fight

One can only hope it eventually does as well as Linux

Alifatisk4• 4 days ago

You don't have to wait to test it. Just now, the Qwen3.8-Max-Preview made its debut on Alibaba’s Token Plan, Qoder, and QoderWork. Be among the very first to try it out.

You can also try it out on Qwen chat, Its free.

57016524005• 3 days ago

in my experience of 1 month daily use, Qwen 3.7 Pro is just unusable. wastes too much time, goes off track, useless stuck loops, cannot debug at all. Deepseek V4 Pro is night-and-day compare to Qwen. actually Qwen models seems the worst SWE experience so far. and it is super expensive compare to Deepseek. cannot delegate anything to it, cannot use it real-time low-level tasks either. totally unusable.

ph4rsikal• 3 days ago

Qwen 3.7 Pro is just unusable. wastes too much time, goes off track, useless stuck loops, cannot debug at all. D

Anthropic should not have bugged their knowledge distillation attacks.

chewz4• 3 days ago

Anthropic should not have bugged their knowledge distillation attacks.

It is like one of Pizzaro's men crying that someone have stolen his precious golden dublons

As Lenin have said - "Loot the looters" (Russian: Грабь награбленное)

chewz4• 3 days ago

From my experience Qwen-3.7-Max is above the Opus level but delivers results much faster. Slightly worse then Fable. Way ahead of Deepseek 4 Pro (in speed and overall comprehension) - which is a workhorse on its own. I am using them all with Claude Code mostly.

Qwen-3.7-Plus is quite OK, good for subagent use. Way better then Sonnet.

Qwen-3.8-Max-Preview seems working just fine for me at the moment - I am playing with is right now but too early to say anything. At 10% of regular price it is a steal so far.

nullbio2• 3 days ago

If by Opus you mean Opus 4 and not Opus 4.8, then sure.

exceptione• 3 days ago

  > At 10% of regular price it is a steal so far.

What price do you see?

Here standard plan has been discounted to $18.00, from $25.00/month.

gigatexal2• 3 days ago

What the difference between your experience and https://news.ycombinator.com/user?id=5701652400? ‘s?

Such diametrically different ones.

amelius• 3 days ago

Can we please include information of what languages we use when making claims like these?

It makes a huge difference if you're writing Javascript/HTML/CSS, Python, or C++/Rust.

Also the application type matters, e.g. user interfaces or scientific computing.

gchamonlive5• 3 days ago

It's useless to talk about models and harnesses without context and method. Depending on how you use the model and what the model is used for, experience may vary drastically. Also, different models with different harnesses require different approaches.

I've been using https://gitlab.com/gabriel.chamon/orisun which is my own simplified methodology, for coding web apps in python and elixir and have been very successful using qwen3.6 27b Q4 locally with help of larger models for architecture, so I get very suspicious when people talk how useless larger models are. They are either using it for a domain that models don't perform well or just not using it right.

big-chungus4• 3 days ago

Qwen3.7 pro is meh, but 3.7 max is a very good model

Demiurge• 3 days ago

Are these different models or different efforts for thinking (internal back and forth review) using the same model?

2Gkashmiri• 3 days ago

Can you tell me more about deepseek?

I paid $2 for deepseek api, put the key in void editor and made a crypto tool in html.

It turned out to be around 67kb. I used sample files in CSV that were a few hundred lines.

It spent around $1.8 in the hour or two or light coding and follow up bugs.

Is it really really this much?

I can't imagine spending a month using it for a day job, it would cost more than the salary so what gives?

I understand the local ai and all that but do cloud providers cost this much?

Earlier I thought "billion tokens" but now not sure

aduwah• 3 days ago

A local AI is not about cost. In fact you will likely pay more for it than with most providers. Just look up the advantages of having access to a technology like this that can be self hosted

57016524005• 3 days ago

so Deepseek 4 Pro cannot go on own sessions for too long.

I delegate small-medium tasks: refactors, summaries, research, writing tests + have very good codebase already + extensive history / architecture / docs / linters. so it picks up and does decent small-medium scope work. it is fast, accurate, cheap. does exactly what I want directly and does not waste time nor tokens.

definitely not "implement me complex greenfield project".

k__• 3 days ago

My 2 weeks with DeepSeek V4:

Pro is ~50% more expensive than Flash.

Both need babysitting.

Plan, split in small tasks, give it docs, types, tests, linter, best practice examples, etc.

Always start a new session when starting a task.

Do regular manual sanity checks, and tell it to find issues in the codebase.

I pay like $1,50 per day for Pro.

3abiton2• 3 days ago

in my experience of 1 month daily use, Qwen 3.7 Pro is just unusable. wastes too much time, goes off track, useless stuck loops, cannot debug at all. Deepseek V4 Pro is night-and-day compare to Qwen. actually Qwen models seems the worst SWE experience so far.

I have used both Qwen3.6-35B and Qwen3.6-27B locally (both Q8 quantized with llama.cpp). I have also used antirez's quant of DS4-flash. They all performed within the same tier, DS4 being a bit more efficient, but they all gave really good results, mainly used for bash scripting, debugging, python and some C++. I am curious what type of applications/langauges failed with Qwen? One thing to note, the chat templates were "broken" for qwen models and had to debug it, there are already effort on this. Tbh, the same with gemma.

Grimblewald• 2 days ago

for usable local layperson applications, I think qwen models are king. However, at hosted/frontier scales I'd agree for code. However for visual comprehension etc. for me qwen is the top of the line, and it isnt even close.

Qwen VL models nail tasks frontier models dont even get close to acceptable on.

ernsheong3• 3 days ago

So are locally-runnable models frozen at Qwen 3.6 now :/

worldsavior2• 3 days ago

Everyone wanted open models that would challenge Opus and Codex, here, you got it.

ernsheong3• 3 days ago

We need better coding models that can run on local hardware, i.e. 128GB VRAM or less

tormeh• 3 days ago

Is qwen 3.6 27b the best model you can run locally at the moment? Not that I have the VRAM for it, but just curious.

cmrdporcupine3• 3 days ago

Gemma4 models are arguably better. Or at least about the same.

ch_sm• 3 days ago

In my experience, yes. A bit more reliable than gemma for me. I mostly use A3B (35B, mix of experts) though, because it‘s faster, and in the same ballpark intelligence wise as the dense 27B, so it’s the sweetspot for me. I want to try cohere‘s mini code model next, but worried the runtimes aren‘t optimized for that yet.

hnfong• 3 days ago

People have been able to run DeepSeek v4 flash with a high spec Mac.

androiddrew3• 3 days ago

I have been running 3.6 27b on a dual AMD r9700 setup using Opencode and Matt Pocock's skills workflow for writing Golang CLIs. It's decent, but won't win any awards on code architecture. I guess you can try to AGENTS.md the deficits but I am just exploring its raw Opencode experience right now. Much slower than an API but still 3x times faster than I can read. Tuning it in with a community chat template and a specific penalty for repeats was the sauce needed to get it to work. I can probably start loop daddying it now over the tickets Matt's flow creates.

So yeah, it's the best local model I've seen. I am going to try the Qwopus 3.6 fine tune soon with the same spec and tickets and compare the output of both.

ernsheong3• 3 days ago

Yes it's between this and Gemma 4 31B which is much slower, but looks like it won't ever get an upgrade. I have to conclude that the MoE variants are unreliable, and MTP sometimes just can't get tricky formatting right.

seanmcdirmid6• 3 days ago

I actually have long discussions with Gemini about this and have wound up download a bunch of different models for different things. There is no best, just fast but worse, slow but better, agentic or not, reasoning or not great at large contexts, better world knowledge, uncensored, etc…. It’s a bit daunting actually since there isn’t really a one size fits all model that you can just use for everything.

atemerev• 3 days ago

The best model you can run locally is Kimi K3, as long as you have the hardware. If "what model I can still run on a something resembling something I can put on desktop without separate electricity and cooling water inputs", then it is probably GLM 5.2 (can be run on e.g. Nvidia DGX Station workstation). As long as you have about $100k-$150k.

schaefer• 3 days ago

I flip flop between qwen 3.6 27b and qwen 3.6 35b 4b active.

But there’s also the quantization of DeepSeek v4 flash called dwarfstar

dofm4• 3 days ago

Maybe, maybe not. Qwen 3.6 27B is literally just three months old. Hard to predict. Maybe it just wasn't worth making a 3.7, and after all, the 27B release was after the Plus release.

try-working5• 3 days ago

in case someone wants to speculate in why chinese labs open source their models: https://try.works/why-chinese-ai-labs-went-open-and-will-rem...

nullbio2• 3 days ago

I predict that no one will use this and everyone will use Kimi K3.

embedding-shape10• 3 days ago

I've been playing around with K3 a bunch, but the verbosity of the reasoning makes complete e2e agent work basically cost the same as other smaller models, and I'm not seeing a huge difference in quality, just a way longer e2e completion time.

sunaookami2• 3 days ago

Same problem with every chinese model currently, they overthink way too much and take too much tokens and time.

jadbox• 3 days ago

What's the price difference?

rubslopes• 3 days ago

Why? Price? If the reason is performance, I've been using non-frontier models for cheap, and they run great for my needs (GLM 5.2, DeepSeek v4 Pro).

rurban• 3 days ago

We'll probably use it, but for images. Qwen is still the best for images

dluan• 3 days ago

waic go brr

antiloper• 3 days ago

Does anyone have the privacy policy of their token plan available? Want to check if they retain/train on inputs/outputs.

moffkalast4• 3 days ago

Lol, lmao even.

Of course they train on literally everything they get their hands on, like everyone else. If you need privacy, that's what local models are for.

adamtaylor_132• 3 days ago

It's a flippant answer to a real question. Anthropic, OpenAI, and even Grok have "Don't train on my data" knobs.

Whether you trust them is different, but there ARE knobs on other hosted AI companies.

smnplk• 3 days ago

Do this giant open-weight models have less active params and could be run on consumer hardware or no ?

adrian_b3• 3 days ago

Any open-weights model that has ever been published can be run on consumer hardware, even on a mini-PC.

The right question is which is the speed that can be achieved on a given hardware and whether it is high enough for the model to be useful.

Until now, the speeds reported for running big LLMs with the weights stored on SSDs have ranged from as low as a token every 10 seconds or so, to as high as a few tokens per second.

With open weights models that you host yourself, you are not constrained to use any single model, because that is the one for which you pay a subscription.

You can use many models, each for whatever it is more suitable. You can use frequently a small model with a high inference speed, but for some tasks you may actually save time with a better model, even if it is much slower.

In my opinion, even at 1 token per second a big model may be useful for some tasks.

apitman2• 3 days ago

I can only imagine what that does to the poor SSD

netdur• 3 days ago

The only problem I had with Qwen, fine tuning on Colab, it takes 31 t/s while Gemma 4 is around 9 t/s, otherwise, one of best local LLM

vitorgrs5• 3 days ago

SVG's pelican https://gist.github.com/vitordelucca/521c2d63c9b852c622e7648...

Made on the website, so not sure if on the API there's more thinking options...

esrauch• 3 days ago

I feel like the pelican test can't be relevant anymore; the whole point was to to something that wouldn't be in the training set at all and now it is?

LatencyKills• 3 days ago

Agree. It was interesting/fun for a bit though.

joegibbs• 3 days ago

What about an armadillo playing a piano? There are so many potential combinations It would say something if the pelican looked great but the armadillo looked terrible

rhdunn3• 3 days ago

A parachuting flamingo? An aardvark driving a bus? It should be easy to randomize the animal and the mode of transport (or vary it with animal playing a sport) to create images not in the training data.

onlyrealcuzzo2• 3 days ago

How about an animated SVG of a pelican doing the Macarena, profile view, spinning to face the camera on the last beats?

rvz2• 3 days ago

I feel like the pelican test can't be relevant anymore;

It never was. The point of this "pelican test" was for performative reasons, or just for attention of the joke.

It is like trying to test whether if an adult elephant could actually climb up a tree and reporting that some elephants are slightly better at doing that than others while also reporting at the same time that they are all bad at tree climbing anyway.

This is an example of testing for the sake of testing. The "pelican test" tests for nothing.

Alifatisk4• 3 days ago

I remember when they released Qwen 3.7 Plus and Max. These models behaved way different from all prior models, it became too verbose. It wrote multiple paragraphs just to answer my prompt instead of the usual concise and direct way responding to me. I didn't like that at all, and I know Gemini also had this behaviour with with the Flash series until I managed to reduce it a bit with personal instructions (in the settings on Gemini website).

I haven't tried Qwen 3.8 Max yet, looking forward to it. My hope is that its way less verbose. Another thing I experience with the Qwen models is that I do not trust their benchmark scores at all. Have anyone played with Qwen 3.8 Max and can share their experience? Which model it come close to? Sonnet 5? GLm-5? DS V4 Pro? Flash? Gemini 3.5 Flash?

cakbeslik• 3 days ago

Using QWEN models since 2.5. I never used the chat properly but as an API I can say they're quite good, especially when you compare with OpenAI models. Cheaper and almost same level. I will try this now also.

sieste• 3 days ago

What is a "credit" and how does it translate to tokens for the different models?

xyzsparetimexyz• 3 days ago

Its the currency of the future.

whynotmaybe• 3 days ago

They use it in Babylon5 (supposedly) in 2260!

ahartmetz• 3 days ago

Since Euro and Dollar values are reasonably close, you can call both of them credits, maybe

sbinnee• 3 days ago

If it offers more than opencode go, the entry plan looks enticing

rhdunn3• 3 days ago

Does anyone know if they intend on releasing open source/weights variants for 3.8 or whether 3.6 was the last model they are/were doing that for?

rolls-reus• 3 days ago

will be releasing weights per their tweet announcing the model https://xcancel.com/Alibaba_Qwen/status/2078759124914098291

Archit3ch• 3 days ago

Obligatory "Does it answer security questions?".

beefsack• 3 days ago

For those trying to get it to work in OpenCode with a Qwen Cloud Token Plan, this is what worked for me. Note that I've just matched Qwen 3.7 Max for the limits as I don't know exactly what they are.

  "provider": {
    "alibaba-token-plan": {
      "models": {
        "qwen3.8-max-preview": {
          "limit": {
            "context": 1048576,
            "output": 65536
          },
          "modalities": {
            "input": [
              "text"
            ],
            "output": [
              "text"
            ]
          },
          "name": "Qwen3.8 Max Preview"
        }
      }
    }
  }

57016524005• 3 days ago

also, be very careful which API endpoint and API Token you use. make sure you use right one (obseve your quota is used up. if you hit right endpoint quota used almost immediately). so that you do not accidentally burn API endpoint tokens (they are expensive, can easily hit 200 USD / 3 days which do not count towards your membership "Credits", if you say purchased it with 200 USD signup bonus in Alibaba Cloud)

cadlernox• 3 days ago

Nice

corv• 3 days ago

Who is behind this site? Is this another frontend to Alibaba or a reseller in Singapore?

jxmorris12• 3 days ago

Why did Qwen stop producing open models? They've gone from building the best open models ~1 year ago to producing like the 10th-best closed models. I don't understand this pivot at all.

Edit: I saw online they do in fact plan to release this openly at some point – x.com/Alibaba_Qwen/status/2078759124914098291

fragmede2• 3 days ago

It's not a pivot, giving away the weights was a marketing strategy that they don't need to keep up with.

InsideOutSanta2• 3 days ago

They've announced that they're releasing the weights for a 2.4T model soon:

https://xcancel.com/Alibaba_Qwen/status/2078759124914098291

mikae12• 3 days ago

https://news.ycombinator.com/item?id=48966120

sinuhe69• 3 days ago

The title is misleading. The link led to a pricing page/token plan and not about the new QWen 3.8 model.

Schiendelman• 3 days ago

The submission link is to the twitter announcement. The body just has a different link to pricing.

MichaelNolan• 3 days ago

If 3.8 max goes open weight, what are the odds they retroactively open weight the earlier releases?

Gecko4072• 3 days ago

What would be the point? Kimi is kind of forcing their hand.

kennywinker• 3 days ago

Well, if nothing else, posterity.

notnullorvoid3• 3 days ago

Always nice to see more open-weights in the heavy model class. I can only hope this trend continues, causing OpenAI and Anthropic to crash and burn.

blfr• 3 days ago

As much as I dislike 'em, this sounds mean spirited. And Alibaba admits in this very tweet that Fable is next level (it is).

Hamuko• 3 days ago

I want as much misfortune as possible to befall OpenAI and Sam Altman after what they did to the memory market.

notnullorvoid3• 3 days ago

I dislike their practices, but the main motivation for hoping they'll crash is that I think their immense overvaluation posses too much economic risk.

yeodev• 3 days ago

Same, you can think of China whatever you want but they're really good at giving big tech a reality check when it comes to AI. We now got a pretty wide range of open-weight models (from DeepSeek and MiMo over to Kimi K3, Qwen 3.8 & GLM-5.2) and I think it's most important that there's a variance not only between quality / intelligence and also cost.

I mean even the cheapest option for Luna is still more expensive than anything DS or MiMo is offering right now and I think a new Ministral model would also hit hard there because we also need some variance in model sources, we can't rely only on the US and China.

glimshe2• 3 days ago

"China" isn't giving anything. These are Chinese companies leveraging their best competitive strategy at the moment: competing on price.

catigula• 3 days ago

Damaging American AI companies just means that China has the lead with closed models.

China isn’t an altruistic state. They’re an aggressor in many fields, economic and otherwise, and this one of them.

Umair_khan2324• 3 days ago

nice

sidcool• 3 days ago

These models are great, but what's the potential use? No small entity can run them.

alex43578• 3 days ago

China encourages/prioritizes their release because it directly competes with American companies closed models.

raised_hand• 3 days ago

interesting, when will this race end?

monster_truck4• 3 days ago

I really like Qwen, even the Q2KP quants of 3.6 27B have genuinely impressive local performance on a 24GB card. It has been good enough that I am happily giving them $60 right now to try this instead of waiting to try a slightly lesser version locally.

Was there ever an explanation for why we never got the weights of 3.7? I would like sourced quotes and not weird/cringe accusative speculation about distillation, or your take on The Big D.

jared0x90• 3 days ago

do you mind sharing your settings? i just picked up an r9700 to start playing with local qwen3.6 27b and your setup sounds promising and efficient on 24gb.

kadoban• 3 days ago

Different quants, but I've been having success with: https://unsloth.ai/docs/models/qwen3.6 I do llama.cpp

monster_truck4• 3 days ago

Which settings exactly do you want?

As far as the model settings go I just follow what's on the card.

I'm using HauhauCS's models, they seem to do a slightly better job with their "P" quants. Especially wrt patching them to eliminate the "doom loops" that will time out the GPU (esp if you have not already given it the extra power budget, set fans to max, and lowered your max clock by about 8% to save yourself a crash/reboot).

Without getting into the weeds, unsloth covers a lot more ground so, when they're good they're great, but I've also had the most problems with them. Basically, don't be afraid to shop around and fuck with sliders.

These days it should almost always be enough to open up LMStudio, set context to max, K/V quant to 16 or 8, and off you go. I'm using a 7900XTX, with 128GB of memory for the cases where things don't fit. The default settings should be fine otherwise.

I don't know anything about the r9700, seems neat. Looks like it might play nicer with the vulkan backend than rocm, and you might have to `set GGML_VK_DISABLE_HOST_VISIBLE_VIDMEM=1`. Perf should be at least as good as what I'm getting (180+ tk/s prompt, >50tk/s response), which imho is just fast enough that I can read it as it reasons and responds.

You can probably get more aggressive with KV quant (any of the 4's) and bump up the batch sizes (4096/1024 vs 2048/512, etc), it's quite situational. These days there generally should not be any serious loss of accuracy from the former. Keeping the rig cool will likely matter more, so turn your music up to hide the fans and set your rig on top of the AC exhaust lol. Oh and I guess make sure ReBAR is enabled and working, adrenaline should let you know if it's not enabled.

dartharva• 3 days ago

I very much appreciate the existence of these free models, but in my experience Qwen has too high of a tendency to confidently give the wrong answer as compared to other frontier models.

mannanj• 3 days ago

It's an interesting time to be alive when your local models are supposedly the pinnacle of what a free nation is capable of, yet the ethicality of the companies is disliked and their models restrict and limit you so much you root for the models from a socialist/communist state. If it wasn't for the effectiveness of propaganda, tribalism and psyops in this scenario my words wouldn't even be controversial and would just be seen as a truthful observation.

Elzair• 3 days ago

Has there been any news on open weighting Qwen Image 2.0 and WAN?

We are spoiled in the LLM segment, but I would love to see an open source competitor to Flux.2, etc.

mchusma• 3 days ago

I counted the other day and there were at least 12 different providers with "better than Opus 4.5 performance" on Artificial Analysis, Opus 4.5 being Anthropic's December release that many say kicked off the latest acceleration. Which is totally insane competition, particularly given how low switching costs. I personally think that Opus 4.5 level performance is sufficient for most apps and usecases, as they get deployed.

margorczynski• 3 days ago

Which is totally insane competition, particularly given how low switching costs

Which is why OAI and Anthropic will most probably push for more governmental control and bans. Without it their whole income model is cooked.

culi6• 3 days ago

I'm not sure if corporations will be willing to allow LLMs to go the way of EVs.

Then again, all of Chinese models are open. And DeepSeek even publishes research papers alongside their models that go in depth into the methodology. I guess there's not much stopping USian companies from copying

wolttam2• 3 days ago

The U.S. has ~350 million people.

Anthropic and OAI can piss and moan all they want - limiting the U.S. to only their models would hurt the U.S. economy in myriad more ways than the failure of a couple of companies that scaled too quickly. If they get that outcome, the rest of the world would simply keep moving forward with access to open models and tokens at pennies on the dollar.

sampton• 3 days ago

This is reminiscent of the operating system wars and browser wars. In the end there will only be 2 models that can survive. 1: give it away for free or 2: locked in with top notch hardware.

maxrumpf• 3 days ago

"compatible" instead of "comparable."

It boggles my mind how you can train a frontier model but not write a tweet without an obvious typo.

segmondy• 3 days ago

let's see your grammatical correct tweet in chinese.

rrhjm53270• 3 days ago

Well, I think Chinese doesn't really have any grammar most of the time.

aloknnikhil• 3 days ago

It's OK because I'm not prompting the person who tweeted for my usecases.

pvorb• 3 days ago

Now you can be sure they write their announcements by hand. Doesn't really matter, does it?

lardosaurusrex2• 3 days ago

at the risk of upsetting a lot of people mentioning this but like are you really surprised?

if youre going to use ai for everything youre gonna start losing your edge as you focus less and less on what youre doing and this isnt me just talking out my ass, like... the front page here is peppered with study after study and blogpost after blogpost about how its overuse can come to the detriment of one's own abilities and skills.

coca cola had the ad with the magical truck that changed its design, shape and amount of tires it had and if nobody noticed that before releasing it then im not sure why anyone might think that the people peddling the LLMs would somehow be immune to this phenomenon

halJordan• 3 days ago

This is the sort of unmitigated pedantry thats the real problem. Everyone makes typos and errors of this nature. Feymann and Hemmingway both did. You deliberately chose this level of error multiple times when you chose to not put an apostrophe in youre and failed to capitalize proper nouns

Come off that high horse

whyenot• 3 days ago

The people training the model are almost certainly not the same ones writing tweets. I don't know why that is mind boggling. We all make typos, at least the humans among us do.

scotty793• 3 days ago

mistakes are at the moment a signature quality of human written texts, enjoy them

Bnjoroge3• 3 days ago

i'll take that 100% over llm slop.

elxr• 2 days ago

Accurate reflection of Alibaba's cloud platform as a whole. I've never seen more misused terminology and strange word choices than when testing out their cloud dashboard.

This is egregious for a model announcement though, I agree with you 100%.

cloudengineer94• 3 days ago

Things are heating up in China.

Looking forward to see what Antrophic and OpenAI does next.

ycui73• 3 days ago

OpenAntrophic and OpenOpenAI ?

kingo553• 3 days ago

Over Dario's dead body.

At least Sam have us a nice model last year with the gpt-oss family.

softwaredoug• 3 days ago

It feels like an inflection point of lost US leadership in technology? A year plus ago you would say while China led in green energy and manufacturing, at least the US was ahead in software - as demonstrated by the state of US AI models.

We could point at a lot of factors on the US side. From political paralysis / head-in-the-sand attitudes towards emerging tech like green energy. To something of disdain for workers that will be impacted by AI (creating a backlash). To education that continues to lag. Add to this so many other self-inflicted economic wounds from the current administration.

I don't know if its nearly as terminal, as say the UK after WW2. The US is still large, wealthy, and resource rich. Yet at a minimum the triumphalism about US leadership after Trump was elected by the tech elite feels silly in retrospect.

Something I also think about is how much stronger The West overall would be if instead of antagonizing allies, there was a single ecosystem working closer together.

pessimizer2• 3 days ago

We could point at a lot of factors on the US side.

It's just lack of antitrust enforcement.

China pours money into tons of different businesses in the same industry and lets them fight it out. The only US business model left now is to shut down (or collude with) competitors and raise prices while cutting costs. All they have to do is cut Congress (and regulators, and individual judges) in. We've financialized everything for the sake of scammers, rather that finance being used for the sake of getting cash to the most productive organizations. We've optimized for corruption.

If we hadn't let the stupidest people in the world buy up everything, and made doing nothing with it the most profitable option, China would have never have blown past us.

The US Supreme Court has explicitly legalized "tipping" politicians. That's the biggest sign of degeneracy that a government could possibly achieve.

https://en.wikipedia.org/wiki/Snyder_v._United_States

2001zhaozhao• 3 days ago

Now there are not one, but two incredibly powerful open LLMs. I think this level of capability makes general prioritization / high level decision making doable with the right harness, and now everyone has hard-to-interrupt access to them (since these are open weights and someone in the world is going to run them). This world is going to get really weird soon, in both good and bad ways...

theplumber• 3 days ago

Let’s better download them fast before Dario is making a scene again!

kingo553• 3 days ago

If he bans them, they'll just show up as torrents. Or we'll download them from Chinese hugging face.

overgard2• 3 days ago

I've been using Qwen 3.6 27B with LMStudio, and I was pleasantly surprised with it, although it was a little slow. I found mtplx last night, and it really wasn't an exaggeration to say that it ran the model 2-3x faster which was super impressive.

I'm trying to move to local models as much as I can, and I'm finding that it's becoming more and more practical. Admittedly this is on a $6000 dollar laptop (M5 Max Macbook with the specs maxxed out), so the hardware is still a bit out of reach for most people (the AI industry isn't exactly helping here..), but I'm getting the impression that the future is going to be smaller models with more focused training running locally. The danger of giving all your data to these cloud providers just seems too big to me, and I think they're going to start charging insane amounts when they need to show a profit.

Aurornis• 3 days ago

Admittedly this is on a $6000 dollar laptop (M5 Max Macbook with the specs maxxed out)

Sadly that laptop is likely $8000 now, after the recent price increase.

overgard2• 2 days ago

Woof, I thought you might be exaggerating but I priced it out and you're right, it's actually way over $8000 now. I hate how the AI industry is making everything more expensive for everyone.

bertili• 3 days ago

Wait.. the Qwen Max models have never been open-weight. But it sure sound like that's what they intend now?

"Qwen3.8 is launching and going open-weight soon! With a massive 2.4T parameters..."

simonw5• 3 days ago

(I can't draw a pelican for this one because Alibaba Cloud have flagged my email address and won't let me pay them for access. So I'm waiting for the open weights release, or for the new model to show up on OpenRouter.)

samxli• 3 days ago

lol the pelican benchmark is basically the only review process I trust at this point. kinda wild that alibaba of all companies is making it hard to give them money tho, you'd think they'd want prominent devs testing their stuff.

openrouter usually picks these up pretty fast, hopefully it shows up there soon. You can also just download the qwen app and do this in the chat interface using their MCP tools for local dev.

timClicks• 3 days ago

It's an interesting example of the metric becoming a target. Simon started drawing pelicans because it wasn't something that existed before.

nextaccountic2• 3 days ago

kinda wild that alibaba of all companies is making it hard to give them money tho

Okay so this is entirely off topic but, are there any options to subscribe to these companies without having a credit card? (I am Brazilian if this makes a difference)

It's kind of wild that a Chinese AI lab will require something like.. a credit card with visa or mastercard label, and very little else.

ahknight• 3 days ago

Did you know you can create more than one email address?

benjaminoakes• 3 days ago

In my opinion, simonw shouldn't have to play those games

formvoltron2• 3 days ago

AI rabble-rouser!

cypressf2• 2 days ago

Here's what qwen 3.8 max preview generated when prompted with "create an svg of a pelican riding a bicycle"

https://cypressf.com/qwen38-pelican.svg

Notably it added a cheeky fish in the beak.

simonw5• 2 days ago

Wow, that one is really good.

simonw5• 2 days ago

What method did you use to generate that one?

I tried using their OpenAI-compatible API and got one that's a lot less fun (no animation) - https://tools.simonwillison.net/markdown-svg-renderer#url=ht...

simonw5• 2 days ago

I managed to get myself a working Qwen Cloud account, here's a pelican. The reasoning trace is pretty fun - three details I liked were "Could add helmet? No." and "Maybe add small bell? no." and "Need maybe add small fish in basket? Not necessary."

https://tools.simonwillison.net/markdown-svg-renderer#url=ht... (scroll to bottom for pelican)

blmayer• 3 days ago

Looking forward for a 20ish billion parameter version

kumanday• 3 days ago

I did some comparisons to Kimi K3... and found the best is combining the two! Model diversity is powerful.

https://trilogyai.substack.com/p/qwen-38-max-benchmark-how-i...

cnhwl2• 3 days ago

I’m a developer from China. So, is this what Hacker News is all about? Whenever a model comes from China, the comments section stops discussing its technical architecture, optimisation points or real-world performance, and instead starts going on about politics, human rights and all that rubbish? To be honest, we Chinese IT professionals possess a genuine geek spirit. That’s why you’re lagging behind in open-source competitions—it’s got nothing to do with politics; you’ve simply lost sight of your original aspirations.

adithyassekhar• 3 days ago

Unfortunately this is a US centric site, for a US based investor, who themselves and probably most commenters as well who are highly paid silicon valley people who have stakes in ai companies on their side of the pond.

Not to mention the political unrest and fear of losing the technological superiority which they once earned but now is trying so hard to hold on to through legally grey monopolistic practices. You dare not outcompete the US in any tech they can sanction.

At some point people who made genuine innovation and wanted to make the world a better and equal place were replaced with capitalists.

Wait for the downvotes.

cnhwl2• 3 days ago

Rankings have never been something I’ve cared about. Just as with the smear campaigns against China, there is far too much noise in the world. We simply hope that the geeks who are genuinely committed to making the world a better place can unite and focus on getting things done, rather than being brainwashed by ideology and social media.

villish6• 3 days ago

Legally gray practices? Such as what?

Old politicians who barely understand how to use e-mail had a panic attack due to Anthropics marketing about Mythos and being told that Chinese models were "attacking" (distillation) US frontier labs to train theirs.

They pull the only lever they could in an attempt to stop it, but again, they don't understand how any of it works. The export ban lasted a few weeks yet everyone keeps crying about it like the US is shady.

In reality the entire world is 100% dependent on the output of US and China for frontier AI. Anyone outside of those two countries will be loud about either closing off access. I personally think it is inevitable that China and the US stops producing frontier level open models when they become too valuable.

throwa3562623• 3 days ago

Yes, and its geting annoying

If you know of another forum with less politics or VC talk and more technical depth please let us know :)

sdoering2• 3 days ago

human rights and all that rubbish

Denying human rights. Classic. Sorry - you instantly lost any respect I could have maybe had for your opinion.

epolanski2• 3 days ago

I think he means rubbish as in completely off topic.

swznd• 3 days ago

he never said anything about denying human rights

gabriel666smith2• 3 days ago

I don't think you're taking the good-faith reading of the sentence's grammar here, which is:

"[When China is mentioned] the comments section stops discussing [technical aspects], and instead starts going on about [non-technical aspects], and all that [stuff which I do not feel is relevant]."

I'm struggling, though, to find a good-faith reading of what you meant by:

"Denying human rights. Classic. Sorry - you instantly lost any respect I could have maybe had for your opinion."

Specifically, "classic" - classic what? "Classic" meaning "a trait you personally believe is more prevalent in people who (might be) citizens of a specific country"?

Hopefully the good-faith interpretation of GP's comment, which you might have unintentionally missed, will help you engage with their opinion with more respect.

I think it's an easy opinion to empathise with personally. I would be enormously frustrated myself, on an individual level, if interesting technical achievements made by companies in my own country were overshadowed by political discussion about the actions of the country overall, when spoken about on technical forums.

This would be especially true for me if I - as an individual - disagreed with politically with my country's leadership. Frustration at a topic of discourse is not mutually exclusive with being in favour of human rights.

I don't think GP has done themselves any favours with the broad statements that close their comment, and is clearly frustrated themselves. That there's mutual frustration is probably a sign that empathy on both sides might lead to a really productive discussion - but focusing on the more substantive points they've made, rather than escalating by misinterpreting the less important aspects of what they commented - is what will lead to that.

midwain• 3 days ago

Denying human rights.

Taking words out of context. Classic. Sorry - you instantly lost any respect I could have maybe had for your opinion.

toasty228• 3 days ago

I don't see anyone talking about Trump, ICE executing people in the streets, the insane level of US politics, the slow build up of the surveillance state, etc when discussing claude or openai

jpleyden982• 2 days ago

Perhaps whenever a US firm, eg. Open AI, Anthropic, Gemini, Grok etc. release a model.

Top comments should be criticising US torture camps in Guantanamo bay or their recent bombing of a girls school in Iran.

Suzuran3• 2 days ago

You are not allowed to discuss human rights until your government stops its unprovoked war on Iran, issues a full repudiation of the policies that led to it, and makes full reparations to the entire rest of the world whose energy prices you have driven up for the benefit of a handful of wealthy inside traders.

Edit: Parent poster wasn't American. I incorrectly assumed their nationality.

tmoertel• 2 days ago

I don't believe the GP is claiming that the idea of human rights is itself rubbish. Rather, the GP is claiming that what is rubbish is the practice of veering into the topic of human rights whenever a new AI model is announced from China.

ksimukka• 3 days ago

What else would you expect from a "majority" of US educated geeks?

Btw, are you using GPUs/TUs at all for local inference? I've been exploring what hardware options exist outside of Nvidia.

mannycalavera42• 3 days ago

instead starts going on about politics, human rights and all that rubbish

look, it's just a matter of priorities ¯ \ _ (ツ) _ / ¯

alex7o• 3 days ago

People in the US have not lost soghts sight of their original aspirations. They are always on track much more than before really. Making as much money as possible in as little time as possible.

rapidaneurism• 3 days ago

The response was similar when the new grock came out.

ghosty141• 3 days ago

humans right and all that rubbish

Ummmm....

I see both technical and political topics being discussed and I think the balance is fine right now. The pricing is also very suspicious such that people raising suspicion about it being subsidized is not that surprising to me.

meowla• 3 days ago

Dude, you're a Chinese developer! I'm also a Chinese developer, nice to meet you.

So do you agree that we developed quite a profile of low-quality software under Chinese techs whose UI and logic are pure not-well-thought-through and even disgusting to use.

Do you really think Alibaba would make their model open-sourced if the model gets to be the best of the world

And how would you comment on the fact that CCP recently restricted high-profile model trainers' freedom of movement by requiring a license whenever they want to leave the country?

They also did the same thing to Manus' Xiao Hong(who is not a model trainer in this case), who looks like a good person

fragmede2• 3 days ago

Welcome! What other sites like v2ex.com are out there on your side of the world?

benterix2• 3 days ago

human rights and all that rubbish

While I understand your point, please think a bit how you formulated the sentence above.

simondotau• 2 days ago

I almost choked on my tea when I read that. I’m glad I’m not the only one who noticed. A shocking turn of phrase, and one that IMHO catastrophically devalues the nominally valid argument underlying that post.

wvlia5• 2 days ago

I think it's well formulated.

eurekin3• 2 days ago

I have a feeling this comment will make history

adornKey2• 2 days ago

In case there is a decent forum about tech somewhere in the Alibaba cloud, I'll be interested. HN has built a certain unhealthy bias in a lot of subjects. But so far other places haven't gained enough traction, yet.

Anonyneko• 2 days ago

It's probably in Chinese, which would cut out most of us here.

woctordho• 1 day ago

There is a forum named Zhihu. AI translation works mostly well to translate contents there into English.

Springtime• 2 days ago

I looked at the comments here but it doesn't appear to be an accurate characterization of the discourse. If anyone has links to such comments shaping the discourse from before the parent posted I'd be interested.

There's one top-level comment that's just an opinion of censorship of Qwen models (which might be argued to be politically adjacent if one assumes the censorship is referring to sensitive political subjects) but no specifics and there was a rebuttal reply anyway, while an unrelated top-level comment has a child reply believing China models are commoditizing AI to unseat US based models (could be argued a whiff of politics but also just competition) but it was still a positive impression.

Then I went to archive.org to check the earliest crawl of this page (prior to parent's post), but nothing different. So I enabled showdead in my settings to check dead-marked posts and there's one post near the bottom talking about how Chinese models are rarely ground-up foundational models, while another dead comment had an angle about it's a geopolitical strategy (neither of which had replies). Ironically some of the dead comments are praising Qwen.

anon5868• 2 days ago

The comment timestamps already show how old the posts are[1]

Putting politics aside, technical chatter is scarce in large‑model discussions here.

Take the K3 release as a case study: out of more than 1,200 comments, how many actually delve into topics like LatentMoE, KDA, or attention residuals? And how many are offering guesses about per‑head Muon or SiTU?

Users interested in this should look at imageboards

[1] https://news.ycombinator.com/item?id=48966211

raxxorraxor• 2 days ago

Perfectly in agreement, but in many places politics were forced into geek spaces. Formerly nobody cared which country you came from.

I think anyone that used open models will have had contact with Qwen. China currently does the most work here, simple as that.

TomK322• 2 days ago

You might want to ignore politics all day long, but rather sooner than later it's politics that will creep into your life and take an interest in whatever things that politicians mis-understand about your little niche.

TomK322• 2 days ago

[flagged]

tomhow• 2 days ago

Please don't reply to anyone like this on HN. The guidelines make it clear we're trying for something better here, and none of us would like it if we were spoken to this way due mainly to our nationality. https://news.ycombinator.com/newsguidelines.html

kwanbix• 2 days ago

human rights and all that rubbish Crazy that you think human rights are rubbish. It clearly paints where china and its population is headed.

orangeboats2• 2 days ago

I think interpreting the word too literally may not result in a productive discussion here. Especially taking into account that the original comment is translated from another language.

All things considered, it seems plenty obvious to me that "rubbish" is used to express the commenter's frustration.

More particularly, their frustration about how the discussion always devolves into non-technical discourse (the actual thing that they are referring to when they say "rubbish", and "human rights" is one example of such discourse) anytime China is mentioned.

They are frustrated because they think this is in direct contrary to the Hacker part of Hacker News, because Hackers should be more interested in the technology.

synergy20• 2 days ago

you probably are overly sensitive and biased, do not forget transformer is the only 'original' for AI so far, the 'original aspirations' nowadays are just incremental improvements. human rights are not rubbish by the way, you might earn high salary now, dare to criticize Mr.Xi at all? you could be in jail and lose everything overnight. without human rights, tech is fragile and mostly harmful in the end.

egorfine• 2 days ago

I have asked local Qwen 3.6 What happened in Tiananmen Square in 1989 and it refuses to answer no matter how I ask. This is why.

On a more serious note, I am immensely grateful for Qwen and I'm rooting for chinese models as opposed to American models - due to the general direction where US models are going.

tracker12• 2 days ago

It's a western site in English where a bulk of users are in the US, which has an adversarial relationship to the US... where many working in IT aren't even allowed to consider Chinese services or products directly (even if what you're getting was half made in China or controlled spaces).

I can't even consider Chinese AI tools for my own use on my own time on my own hardware without potentially risking firing or worse. Yeah, it's cool on a technical level.. but the "open" part of this and K3, etc. is of limited use depending on how/where you can actually use it yourself. Trust is another issue entirely.

fc417fc8024• 2 days ago

the "open" part of this and K3, etc. is of limited use depending on how/where you can actually use it yourself.

... from servers owned and operated in the US by US providers? Go check out the list of providers on ex openrouter for ex the latest deepseek model. Some of them even fall under the zero data retention agreement.

andriy_koval• 2 days ago

That’s why you’re lagging behind in open-source competitions—it’s got nothing to do with politics

there are plenty oss geeks in US, but they don't have access to similar compute, alegelly subsidized by Chinese government.

nullc3• 2 days ago

That’s why you’re lagging behind in open-source competitions

the lag is because closed ai orgs are sucking up most of the talent and monopolizing resources.

Not that your criticism of HN comments aren't deserved-- but the people commenting here are not usually people who would be doing something useful otherwise.

sam345• 2 days ago

Oh please. Yes it's what it is about. China stealing western technology and flooding the market with underpriced stolen ip to bolster dependency.

tesnorindian2• 2 days ago

Chinese open weights models have liberated the world from Western exploitation. Great respect for Chinese students especially those who returned from foreign countries after their graduation and their research on AI.

alexgoodhart• 2 days ago

The United States and China are both extremely powerful governments with dispositions towards their own citizens and the outside world that are unequivocally despicable. I don’t think that the discussion of Chinese tech devolves but you’re right that China’s geopolitics flow into the convo easily while America’s does not.

This website isn’t exactly a forum for niche experts of every type. It seems more like a group organized around respect for the people who actually can speak at length and specificity about technology, while ultimately most of us are just well informed and tinker-level consumers.

senordevnyc• 2 days ago

starts going on about politics, human rights and all that rubbish? To be honest, we Chinese IT professionals possess a genuine geek spirit.

An incredible amount of damage has been done in the world by "geeks" who lack a sense of care for ethics, human rights, "and all that rubbish".

The_resa• 3 days ago

Open Source Era is comming

phs318u• 3 days ago

Off topic, but comment threads like this, where the early comments rabbit hole into something I didn’t come to read, really make me wish HN had collapsible comment controls. My phone screen nearly melted from the insane amount of swiping/scrolling to get to the part of the discussion about the actual LLM.

boutell• 3 days ago

I'm making good use of the next button on original replies, but it's true that walking back through the parent button is a long walk sometimes.

Schlagbohrer• 3 days ago

I have been using the docling software for PDF and XLSX reading/viewing/comprehension by my local qwen3.6. Claude was telling me that docling includes within it a very small LLM model just to assist with what is basically "super OCR". We are definitely in this era of ultra-mega-huge frontier models and super-tiny-micro models, all finding their uses.

On an aside I tested docling by giving it the Ronin TTRPG rulebook PDF and it did an astounding job of converting it all into .txt and .md. Given the wild graphic design of the book that's pretty astounding. Next we'll see if it can read the other Borg books like Mörk Borg and Cy Borg, which are also famous for their messy information.

jchoksi• 2 days ago

An alternative to Docling is Xberg: https://github.com/xberg-io/xberg

"A polyglot document intelligence framework with a Rust core. Extract text, metadata, images, and structured information from PDFs, Office documents, images, and 97+ formats. Available for Rust, Python, Ruby, Java, Go, PHP, Elixir, C#, R, C, TypeScript (Node/Bun/Wasm/Deno)- or use via CLI, REST API, or MCP server."

Amongst other things, you can use it with Open Webui: https://docs.xberg.io/integrations/openwebui/#choosing-an-en...

fyi, they are releasing their v1.0.0 release shortly and until then you will need to a release candidate docker image tag: https://github.com/xberg-io/xberg/issues/1192

As a layman, I would say Xberg is better than Docling because its one engine for everything while the former feels like a bunch of scripts and glue logic.

EchooAI• 2 days ago

Without a doubt, the Chinese models are going to overtake the American ones. In my opinion, this will happen this year!

HerrlichDigital• 2 days ago

I used Qwen before, and I'm pretty curious what the new version is all about.

niko323• 2 days ago

Keep up the great work Qwen. You give us options.

Helldez• 2 days ago

I tried it and it's incredible

elaz48• 2 days ago

the pricing is this competitive

zftnb666• 2 days ago

Qwen's models keep getting better. Same problem as the other Chinese models though — international payment is painful. If anyone wants to try Qwen or DeepSeek V4 without the payment headache, api-hub.cc supports both with credit cards.

Qwen 3.8

Related Stories