Nativ: Run frontier open models locally on your Mac

https://blaizzy.github.io/nativ/

by aratahikaru5 • 2 days ago

374 127 2 days ago

blaizzy.github.io

rvz• 2 days ago

Looks like my call [0] for more competitors to Ollama has been answered.

We need more like this as well as llama.app, which also has a native mac app.

[0] https://news.ycombinator.com/item?id=48968898

laughingcurve• 2 days ago

Agreed. Just based on this not being Ollama, so I will give it a try.

0gs• 2 days ago

my thing is kind of an Ollama competitor (surrogate?) too. more for prose/text planning, not so much for coding, at least the harness, but i'm sure someone could set it up to do that: github.com/0gsd/enough

iAMkenough• 2 days ago

How does this compare to LM Studio Bionic?

asqueella• 2 days ago

Bionic is an agent; this appears to be an open-source competitor to lmstudio. The initial commit is just a few hours ago though, so …

calumcl• 2 days ago

The maintainer works on mlx-vlm so he does have pedigree in the scene, I'm don't know if this is just going to end up as unmaintained slop. I haven't tried this but I would personally just recommend oMLX for a currently more complete and fleshed out package - loads of features, provides the same MLX support and changelogs + commits are actually detailed.

crefiz• 2 days ago

Why do ppl downvote this comment? HN makes no sense sometimes...

SafeFatNoob• 2 days ago

so now we have this, https://pypi.org/project/rapid-mlx/, https://mtplx.com, and the oldest I could find at https://omlx.ai.

philips• 2 days ago

There is also https://github.com/ml-explore/mlx-swift-lm which is what https://jan.ai uses.

moontear2• 2 days ago

Is „frontier“ overused? I thought frontier models were the best-of-the-best such as Fable right now. I assume you can’t host these models yourself since you would need many GB of RAM and expensive GPU of is my thinking of „frontier models“ wrong?

44za12• 2 days ago

+1 came here to say this, I opened the link expecting some technical breakthrough. Misleading click bait title.

bnfcl• 2 days ago

Had the same though. The gap between open-source and the frontier is closing in, especially with Kimi K3, but that is like >2T parameters. The Gemma 4 and other models you can actually run on an average Mac, is not in the same league.

zeckalpha• 2 days ago

The frontier is a curve. https://en.wikipedia.org/wiki/Pareto_front

IshKebab3• 2 days ago

That's not what people are normally referring to when they say "frontier models". It means the most capable models full stop. Not the most capable that you can run locally.

kube-system• 2 days ago

That is one sense, but it is also use to refer to the frontier of capability. Especially in this context.

dragonwriter• 2 days ago

The frontier is a multidimensional space defined by the “best” combination of traits a model can have in multiple dimensions: parameter size [smaller is better], various task metrics, and relative token generation speed on like hardware [faster is better], and active memory requirements [smaller is better] are all possible dimensions, and a model on the frontier of the current options space is one where getting better on one of those measures cannot be done without getting worse on at least one of the others.

dofm5• 2 days ago

I don't know that I agree with this specific use of frontier because it is confusable as you say.

But (off on a tangent) I do think that there are multiple frontiers generally — and I also think the open weights, small local model frontier is by far the most important and exciting one.

I keep mucking about with what Gemma 4 12B can do and every time I do I find myself thinking that all the energies in the AI world are going in entirely the wrong direction, because it is small, clever, efficient and remarkable.

If all of that research money were to be spent on improving AI models that fit inside a 16GB RAM machine with unified memory, I think really important progress could be made.

I enjoy using the Qwen 3.6 models (and BottleCap's new fine tune of the 27B) but the small Gemma 4 models are impressive in a way that I think is going quite unreported.

So while I don't think this website should use the word "frontier" here, even referring to Qwen 3.6 27B which is weirdly close, I think it could.

lukevp• 2 days ago

I’ve got a 32 gig m1 MacBook Pro. how would I go about trying Gemma like you mentioned? Would it run at an acceptable speed, and what could I do? Coding?

lucideer• 2 days ago

It might be overused elsewhere but I don't think its use here is inappropriate given it's qualified: it doesn't say "frontier models", it says "frontier open models".

oersted2• 2 days ago

I was confused too, but I believe that it refers to the Pareto Frontier: the best set of solutions to a multi-objective problem.

Look it up, it’s a bit difficult to explain concisely in words but it is intuitive visually.

If we are thinking of intelligence and price, a model will be in the Pareto Frontier if there’s no cheaper model of the same or higher intelligence. Or if there’s no more intelligent model for that price or lower.

EDIT: See this chart from Artificial Analysis: https://artificialanalysis.ai/#intelligence-comparison-tabs

So for example DeepSeek V4 Pro can be considered a frontier model because there's no cheaper model that is as intelligent.

For any solution in the Pareto Frontier, there no "no-brainer" alternative, in the sense that there's no other option that is better in some way without giving up something else. It's the best of its "weight-class".

andsoitis3• 2 days ago

The subtitle on their website is: "Nativ puts frontier intelligence on your desk".

Also, the title is "Run AI models locally on your Mac," not "Run frontier open models locally on your Mac."

moontear2• 2 days ago

Very interesting rabbit hole for today. Thanks for mentioning Pareto frontier!

jboss102• 2 days ago

I think this is one of the easiest to understand.

https://arena.ai/leaderboard/text/pareto

avazhi• 1 day ago

That’s not what frontier means.

There are two frontier labs - OpenAI and Anthropic, and maybe 4 frontier models currently. That term refers to model capacity and in general represents a model’s capability relative to the rest of the industry irrespective of price. That’s why even the Qwen announcement carries language about it being ‘near frontier’. Same with Kimi.

Whether a model is frontier or not has everything to do with capability and nothing to do with price.

jdthedisciple• 2 days ago

No, your thinking is 100% correct. It's called clickbait.

dlandis• 2 days ago

Advice: remove all slop and fluff from the website such as "Everything you need. Nothing you don’t."

Just state the information you want to communicate in the plainest and most straightforward way possible.

vitally3643• 2 days ago

Something nobody needs is pointless hot air like "Everything you need. Nothing you don't."

thejazzman2• 2 days ago

You’d be surprised how hard this actually is. I spent 3 days iterating on a marketing site, where I had very explicit / “well written” copy, and it would just repeatedly rewrite it back to the most awful slop. Over and over again! Ended up adding various AGENTS rules telling it to leave the copy alone

jmpz• 2 days ago

Definitely easier and better than just writing the copy..

satvikpendem2• 2 days ago

Just...write it yourself.

anentropic• 2 days ago

Try https://github.com/blader/humanizer agent skill

para_parolu• 2 days ago

I don’t think anyone ever looked at result. Website if full of overflow bugs. It’s pure slop.

sajithdilshan2• 2 days ago

true, it doesn't render properly on mobile. Although I quite like the design. It's a new AI design guide (font/color choice) I haven't seen in in other AI designed pages

sebmellen• 2 days ago

That phrase itself is such an astounding performative contradiction [0].

0: https://en.wikipedia.org/wiki/Performative_contradiction

isomorphic• 2 days ago

This looks like the Prism folks, who are making binary/ternary versions of popular edge models, so that those models will fit on constrained devices like phones. E.g., their Bonsai model derived from Qwen:

https://news.ycombinator.com/item?id=48910545

Perhaps they got tired of LM Studio, etc., not being able to run their models properly.

dofm5• 2 days ago

Is it? The developer page for the github repo suggests he works at Arcee. Maybe he moved?

(I initially thought the same because of the website appearance)

D13Fd2• 2 days ago

I'm surprised that their home page basically acts as if LM Studio and others don't already do this. It's not clear what the difference is from a glance.

It also omits Open WebUI. I've been running Deepseek V4 Flash locally on my Macbook Pro for weeks using Open WebUI + DS4.

lylejohnson• 2 days ago

I was wondering the same and assuming I'd overlooked something.

kzrdude2• 2 days ago

LM Studio seems to do the same thing, except that LM Studio is not open source. So they have a point, they do something more.

moostii2• 2 days ago

LM studio is closed source software built ON TOP OF code released by the author of Nativ.

a3w• 2 days ago

jan.ai is the F/LOSS alternative to it. If you do want a gui.

wmf2• 2 days ago

"The other “local AI” apps you’ve heard of? They’re proprietary shells built on top of open-source engines they don’t own."

This is a roundabout way of addressing LM Studio.

syntaxing• 2 days ago

What spec is your macbook? I want to run Deepseek V4 Flash but its too slow for agents on my Strix Halo.

D13Fd2• 1 day ago

It’s a maxed out current-gen MacBook Pro w/ 128gb ram. I lucked out in getting it before they upped the price.

It runs really well and fast enough for what I need, and it works great, but it’s slower than Claude. I use it for projects where I’m not allowed to use third party AI for legal reasons.

satvikpendem2• 2 days ago

Only interesting thing about this vibe coded runner is the MLX support, as that's still annoying to use in other ones, most still use GGUFs. Unsloth Studio which is an OSS runner I use is still in progress with MLX support although it's still a ways away.

sajithdilshan2• 2 days ago

Has anyone found a model that can run on a normal macbook? I have an M3 Pro with 18GB of memory and whenever I try to run even a basic model the fans goes off and the mac starts to get heated up and becomes so laggy.

sixtyj• 2 days ago

Tbh, the only reason to run a model locally is when you want to be completely safe.

From productivity point of view, it doesn’t make sense to have any notebook running a local LLM.

We have one life. We should spend it wisely.

mft_• 2 days ago

There is a Gemma 4 model with 12B parameters which might be worth trying. e.g. https://huggingface.co/mlx-community/gemma-4-12B-it-qat-4bit

That said, your computer will still get hot!

Daunk2• 2 days ago

I just run Gemma4 12B MLX via Ollama and it's been doing fantastic work!

kzrdude2• 2 days ago

If you go small enough it should be no problem. For example Gemma 4 E4B in Q6 or Q4 quantization should run well on your laptop. It shouldn't be too taxing, but would still want to eat 7-9 GB of VRAM or so.

Now that model is mostly useful for writing or chatting.

fl0id• 2 days ago

heating up is normal, that cannot be avoided. it should become laggy, but you just have very little RAM (I assume 16 GB?) So most models are too big with other stuff running.

b3ing• 2 days ago

You need more RAM, plus the models take up a lot of space. 32gb min but I’d recommend 48/64gb, you won’t get close to frontier but it’s still fun to play with, images are very good

c4pt0r• 2 days ago

I'm a bit curious why not running DeepSeek V4 on top of https://github.com/antirez/ds4. I think the results could be really good.

Archit3ch• 2 days ago

I assume Nativ doesn't support SSD streaming like DwarfStar.

erikkahler• 2 days ago

To clarify, this MIT-licensed app is from the very same dev, 'Prince Canuma', who maintains the popular MLX-VLM library (https://github.com/Blaizzy/mlx-vlm). MLX-VLM is a long-time dependency of the excellent LM Studio and others because it can provide faster inference on Apple devices than llama.cpp. Historically, MLX is a smaller community than CUDA, but has some of the fastest updates upon the release of new models, particularly in models with modalities beyond text-in, text-out (vision, STT, TTS, video gen). See also (https://github.com/Blaizzy/mlx-audio-swift). Would be totally unsurprised if those modalities and models get integrated into this UI.

Possibly vibe-coded landing page notwithstanding, the app is mostly written in Swift language. That suggests it will be easy to port this inference stack to iPad and iPhone.

trollbridge• 2 days ago

Thanks for that. My first question was “What does this do that Unsloth doesn’t?”

recursivegirth• 2 days ago

LM Studio is trash on Windows / Linux... guess that makes sense...

Forgeties79• 1 day ago

What makes it trash on Linux? I’m a pretty casual user myself so my guess is I haven’t bumped against these limitations. I also don’t have high expectations as I’m running it on a 9070 lol

rahimnathwani2• 2 days ago

  Would be totally unsurprised if those modalities and models get integrated into this UI.

Yup, the GitHub repo says:

  Support for dedicated audio-only and image-generation-only models is coming soon.

Prince Canuma is super-responsive on X and GitHub issues, and I use mlx-audio almost daily with mlx-community/Qwen3-TTS-12Hz-1.7B-Base-bf16 (for voice cloning).

carterschonwald• 2 days ago

its a wrapper around mlx, so thats gonna be the portability bottleneck

simonw• 2 days ago

I was so excited when I saw "blaizzy" in the domain, because Prince Canuma's work around MLX has been of such uniquely high quality.

justatdotin• 2 days ago

Hi Simon, sorry to spam your comments but 7 months ago you asked me for a media report to back a claim I made and this week it finally arrived. perhaps a little late to be very useful to anyone who maintains that gating function, but here nonetheless:

https://ntindependent.com.au/scientist-says-ministers-pole-f...

walrus01• 2 days ago

I would also note that for people who want to download these models, you can find MLX versions of just about everything popular on huggingface these days. For instance go look at the "main" page for Qwen 3.6 35B-A3B and then follow the link to quantizations, and pick one of the more popular/reputable MLX variants.

https://huggingface.co/Qwen/Qwen3.6-35B-A3B

sgt101• 1 day ago

converting a model to MLX is trivially easy as well - just a couple of lines of code and a few minutes wait.

I do find that it reliably crashes on the first try, but then it picks up from where it finishes and completes the conversion reliably.

layla5alive• 2 days ago

Prince CacheAsNUMA? Nice :)

Der_Einzige• 2 days ago

Switching to mlx-vlm is basically harmful since it (like vllm and sglang) have such garbage support for modern samplers. To be clear, I am one of the authors on the min_p paper, and if min_p is the best you have (when llamacpp supports the far superior top-n-sigma), than I have no reason to switch even if you are somehow faster.

(https://arxiv.org/abs/2411.07641)

And if you do care to support modern samplers, you can start with the following:

1. https://arxiv.org/abs/2509.23234

2. https://arxiv.org/abs/2509.02510

3. https://arxiv.org/abs/2604.11012

woadwarrior012• 1 day ago

Thanks for bringing this up. You're 100% right. But most people, even technical ones are oblivious to how much of a difference modern samplers and higher quality quantization algorithms make for on-device LLM inference and are stuck with good old top-p, top-k samplers and RTN quantization.

TBF mlx-vlm does support min-p sampling, but none of the other modern samplers that you list. Ollama and LM Studio are even worse with only top-p and top-k samplers.

jarek83• 2 days ago

Ok, so we're at the point that even design are just one-shot by AI. This is exact look and feel any time I ask it to present a HTML doc about anything.

moostii2• 2 days ago

The webpage isn't the point. The software the webpage refers to is the point.

woadwarrior012• 2 days ago

Ironic that the app is named Nativ(e) and yet bundles a full Python runtime. Nonetheless, still less bloated than LM Studio, which bundles a full Python runtime and electron.js (which in turn bundles a whole browser runtime).

bigyabai• 2 days ago

A lot of macOS apps are statically-linked, even interpreted programs. It's still a native app for going that route.

Daunk2• 2 days ago

Is Gemma 4 E2B actually "usable"? I've been running Gemma 4 12B and it handles everything very well! But the second I've moved down to E4B it's been unable to perform the simplest of tasks. So I can't even imagine how E2B would do... Or am I doing something wrong?

JosNun• 2 days ago

Genuinely curious: what are people using these smaller local models for? They are getting decently capable, but they are still small enough that I don't trust them for "real" work outside of a handful of fun toy projects.

Are people actually using them in coding agents? Or are they mostly using them for other things?

teaearlgraycold• 2 days ago

They're great at helping me look up web dev stuff when I don't have internet access.

kgeist• 2 days ago

We've shipped some code generated by Qwen3.6 27B to production (under OpenCode). It lacks the breadth of knowledge of models like Opus, but if a change is fully inferable from the prompt and the surrounding code, it works very well. It won't be able to write something from scratch that requires niche knowledge (say, a performant inference engine tailored to Blackwell GPUs), but if it's just a PR adding a new use case to an existing project (which is usually just "load from the DB, do some invariant checks, modify the entities, store them back"), it works as well as Sonnet (provided you have the correct configuration, like recommended temperature and top-p settings, the model isn't over-quantized, you have at least 150k tokens of context available, etc.).

jiqiren• 2 days ago

there is plenty of grunt work these smaller models can do. update dependencies, fix merge conflicts, write --help, markdown, or readme files for existing code. etc.

sometimes they fail but undo is just a "git restore" or if automated, rejecting a PR and having a better model take a crack at it.

mips_avatar4• 2 days ago

Qwen35ba3b can do a huge amount of data cleaning work on pretty modest hardware. Already have run about 100 billion tokens on it using 2x3090 gpus.

febed2• 2 days ago

What exactly do you mean by data cleaning

noja2• 2 days ago

are you using it differently to OpenRefine?

netghost• 2 days ago

I don't use them as coding agents, but they can be very useful for things like text transformation, summarizing, or text extraction.

That said, if you have a subscription to a paid model already, you're not necessarily winning out on anything except perhaps privacy, which isn't nothing.

efficax• 1 day ago

I use a gemma4 model locally to extract content from messages to a personal agent I'm building for its memory graph (to break the message up into the topic, source (assistant or owner), facts, entities, etc. in the message content (all getting thrown into a magma-esque graph using NLEmbeddings for memory search). This is for a custom personal agent that targets deepseek-v4 flash. The local model is too slow in my setup for a chat agent, but for memory extraction it works pretty well, saving API usage on every chat turn.

shitcoder• 2 days ago

Looking forward to giving this a try. I have tried MLX using Rapid MLX however the LLM (Qwen) would always have hiccups and get stuck repeating itself.

Moving onto llama.cpp I was able to get faster tokens with MTP and a more reliable llm.

I wonder what other people's experiences are using MLX vs llama.cpp

dofm5• 2 days ago

FWIW on my M1 Max I have not really seen any advantage at all from MLX.

I am fully prepared to believe the benefits accrue more to the M3 and up (because of changes to the Apple Neural Engine).

But with the models I've tested, unless I am missing something, the performance of GGUFs in llama.cpp has been better in some cases.

I still have not had results from Gemma 4's MTP be really worth it, to be honest; but with the Qwen 3.6 MoE it is measurable. Maybe with newer kit it is more meaningful.

(There is every chance that the above is not the experience of anyone who really deeply knows what they are doing; it feels like I am a perpetual novice at this stuff)

regexorcist• 2 days ago

Same here. Tried MLX twice at different times after reading the claims here but it always does considerably worse for me than llamacpp.

TechSquidTV• 2 days ago

The server wont start for me. "ERROR: Application startup failed. Exiting." "mlx-vlm-server stopped with status 3"

jsomedon• 2 days ago

I can't really tell why would I use this over like lm studio, jan.ai and such?

flyingcapabara• 2 days ago

If you are on Linux check out Box, it runs models locally , has img gen etc and many more features, I will be releasing the source soon just polishing out last bug's, help porting to other distributions is welcomed

Github.com/jegly/b0x

saagarjha• 2 days ago

I'm a little confused why the page lists "UNIVERSAL · APPLE SILICON (M1+)". This seems like an oxymoron

khurs• 2 days ago

Flagged this, as it is not 'frontier models' in the title of the linked page so not sure why has been titled that way here.

It's a further way to run mlx models

mlx are apple specific format for M3 or later CPUs, and some benchmarks show mlx are not always better than just running generic ones.

noja2• 2 days ago

Can you add support to download gated models?

shireboy• 2 days ago

What is the “middlest” Mac one could get for this? I’m in the market but keep going back and forth between a 64gb m5 pro or “lower end m5 air and screw it I’ll just pay for cloud tokens”. At current prices the 2-3k diff to try to run something local that isn’t as powerful could buy a lot of tokens.

froobius• 2 days ago

You can run a lot of these on e.g. M1 max 64gb

Nekorosu• 1 day ago

I really don't like the marketing texts. "Why we’re open source when nobody else is." I'm using oMLX which is open source and seems to be doing everything Nativ offers. I'd rather see the comparison with existing "non-existing" open source competitors.

giancarlostoro• 1 day ago

What they likely mean is, why options like LM Studio are not open source.

kmike84• 1 day ago

omlx is quite similar to LM Studio, so there are "options like LM Studio" which are open source

diimdeep• 1 day ago

I am still rocking Sequoia and this targets Tahoe purely from UI constraints, no like.

cootsnuck• 1 day ago

So should I ditch ollama for this?

mfro• 1 day ago

I don't love that this starts an API server on launch and has no option to disable it...

n8henrie• 1 day ago

Any reason to use this over omlx?

reagle• 1 day ago

That's what I wondered.

themihai• 1 day ago

What’s the BS with “Open models from teams we trust.”?

m3kw9• 1 day ago

The site design itself is likely one shotted. The italic fonts is nauseating

jwr2• 1 day ago

The word "frontier" is like "load-bearing" at this point (Claude Code users will know what I mean). I wish we could stop using it. Especially as this does not, in fact, run the leading/top models locally on your Mac.

jwr2• 1 day ago

How is it better than, say, LM Studio? And does it run MTP models (which from what I understand are GGUF)?

it• 5 hours ago

I ran it with Qwen3-8B, and it responded to all input in the chat with

Could not connect to the server.

Nativ: Run frontier open models locally on your Mac

Related Stories