[dwa3592]

I have a prediction. By the mid of 2027, we will have >200B MoE models running on basic consumer hardware.

I am running Qwen3.6-35B-A3B locally on my 16GB mac with 7-9 tokens/second. Link - https://github.com/deepanwadhwa/samosa-chat

This is a GPT4 level model running locally with a decent speed on a 16gb ram macbook air.

[smeeth]

Probably won't have to wait that long. Prism released Bonsai 27B (https://huggingface.co/prism-ml/Ternary-Bonsai-27B-mlx-2bit) as a ternary model a few days ago, its just ~7GB and runs at 44+ t/sec on an m4 max laptop. That's already in the ballpark of active parameter count of most 200B+ models, so we will get a model like this whenever Prism feels like releasing one.

It is debatable if we will actually need that many parameters though, since recursive nets like HRM (https://huggingface.co/sapientinc/HRM-Text-1B) don't need to parametrize as heavily.

[acchow]

We're too easily conflating parameter count with capability. That Bonsai 27B you're running is at 2-bit quantization. Is it really better than the best 10-18B models?

[smeeth]

1. yes. https://www.alphaxiv.org/abs/2607.bonsai-27b table 14 shows that bonsai retains roughly 95% of the fp16 27b model's average performance and outperforms post-training quantization at a similar bit width. it doesn't directly compare against every top 10-18b model, but it is clearly still performing like a large model.

2. quantization != native low precision training. a model trained in native ternary should generally outperform a full-precision model quantized after the fact.

even if a ternary model only retains 90-95% of the performance of its fp16 equivalent, who cares? if a 200b ternary model retains most of the capability of the 200b fp16 model while using a fraction of the memory and bandwidth, it can be substantially less efficient per parameter and still dominate a smaller fp16 model under the same hardware budget.

[Aurornis]

> yes. https://www.alphaxiv.org/abs/2607.bonsai-27b table 14 shows that bonsai retains roughly 95% of the fp16 27b model's average performance

I know that's what the paper says the benchmarks say, but these models feel significantly worse than the base model when you start using them for real tasks.

Even the Q4 quant which they put in between their Bonsai models and the FP16 in the benchmarks has a tendency to go into doom loops and get lost compared to even Q5 or Q6.

I don't know how much of this is due to benchmaxxing (putting the benchmarks into the post-training loop) or cherry picking benchmarks to look good. If you spend a lot of time using local models you learn to take vendor provided benchmarks with a huge heap of doubt. Everything looks amazing in the benchmarks these days.

[benjiro29]

Models are not like a fps counter where losing a few percentage has no actual impact. These percentages may be the difference between a model that writes code and one that goes into loops, and think rm -rf . is a good idea.

There is a reason why most models try to stay in the FP4 or higher range, because the reduced accuracy can have major consequences.

You are better off with a 8b FP4+ model then a 27b Q2 model.

[smeeth]

You're conflating two concepts: native bit width and post-training quantization.

Consider two models: one is 16B and trained natively in 2 bits; one is 8B and trained natively in FP4. These models have the same total number of weight bits, but one has twice as many parameters. There is no real reason the FP4 one should be better just because it's FP4. It might be, but that is an empirical question, not a general rule.

Post-training quantization is another thing entirely. Taking a model trained at higher precision and forcing it down to 2 bits is going to hurt performance, often very badly. But this was never my point.

[busfahrer]

has anyone been able to get bonsai models to run at a proper speed in a x86 windows cpu-only setting, using llama.cpp?

[nkozyra]

No.

But do you need to run every small problem through a 10B-30B model?

We're smashing ants with hammers most of the time. We're asking frontier Opus/Fable models to classify text and build frontend code.

Once we start dissecting these problems into smaller discreet tasks and having the big reasoning models do the tough stuff, we suddenly have an economical system. Not for the company hoping for a big IPO, but for the end user.

[edg5000]

I do about 10 google search queries for every 1 opus/gpt prompt. For google, I don't actually open pages anymore 9 out 10 times; I rely on the AI summary. It's fast and accurate; the trick is that you learn where the boundary is of what you can ask it. Querying information the small model is great at.

Then there might be slow, batch tasks. I can see myself getting 1T of slow RAM one day (in a few years?) and having a slow onsite GLM5.2 doing batch jobs that would be wasteful of my subscription limits, plus sensitive but boring things, such as bookeeping and general admin.

I'd like to to read all my email and al quarterly reporting. But that would have to be a good local model, probably a model simmilar to whatever google search uses, which seems just correct unless you throw serious challenges at it.

[ubercore]

The big hammers buy you more confidence and need less supervision. In pure task execution you _might_ smash the ant with a small surgical hammer, but if you absolutely need it smashed, that's when people still reach for the big hammer. It buys more confidence.

[mdp2021]

> do you need to run every small problem through a 10B-30B model? ... We're asking frontier Opus/Fable models to classify text

Actually probably yes: text analysis (magazine articles) by LLMs in the ~30b .. ~120b range failed miserably (and also randomly - the rare cases of proper interpretation occurred among the failure cases) with the main public models of around one year ago, tried extensively.

So, yes, you can employ an ~80IQ only if you will expect the related quality.

[acchow]

I don’t think you need a 10-30b model for most smartphone use cases.

But I meant to counter gp’s claim that “I can run a 27b model on an iPhone” is kind of pointless and disingenuous. Yes I’m sure someone will come up with a way to run a “27b model” at 0.1 bit quantization on an Apple Watch pretty soon misses the whole point of saying a model is “27b” in capability.

Achieving a parameter count is not the point. And is almost meaningless

[yieldcrv]

Its somewhat good, the prism team’s webgpu demo gave it a couple dozen “kernels” written in Fable 5 and it calls them for almost everything procedural

I feel like these things are experiencing convergent evolution to be like biological brains. The large parameters are merely potentially large parameters and they keep having more and more and smaller active layers, which are themselves quantized down. This is seems analogous to the chemical spiking of neurons and inactive layers of a brain in power and efficiency.

[refulgentis]

There’s a good eval floating around somewhere and tl;dr they’re awesome but the benchmarks are cooked, you’re better off with Qwen 8B Q4 than 27B 1b or ternary.

Thanks for being skeptical, I maintain a llama.cpp-based client and it’s frustrating how high expectations are for local AI bc the median effort level means people mostly assemble their expectations and understanding via marketing soundbites

[flockonus]

There is also the 1bit version @ 3.9 GB that retains 90% of the intelligence - quite a feat!

https://huggingface.co/prism-ml/Bonsai-27B-mlx-1bit

[dwa3592]

agreed!! in my heart i really wanted to say by the end of 2026 but wanted to add some wiggle room in case they start to ban open source AI development.

[dTal]

>in case they start to ban open source AI development

Is that likely, do you think?

[submain]

NVidia seems to want them. They have a lot of lobbying power.

[dwa3592]

dario has been saying open source models are dangerous. who knows who is listening to him.

[mdp2021]

The one time in which I saw Juergen Schmidhuber in perfect nervous control, "coolness" they may say westward, was when he replied to one member of the audience, "The same observation was made when they invented fire: oh, it's dangerous. But in the end, now it's here (shrugh)".

There is a proposal in the USA to restrict LLM access. This will only have us depend more and more on open source models and their providers. And cause a drain of research in those areas in which it will be impeded.

[reidrac]

> on a decent speed

But you said 7-9 tokens/second, that's not a decent speed. I'm not an expert by all means but in my local experiments, less than 12 to 16 tps is too slow.

[LeifCarrotson]

I think that any workflow that requires the user to stare at the tokens being generated live is using it wrong. Delegate, don't stare!

https://mikeveerman.github.io/tokenspeed/?rate=10&mode=text

You think of an idea that you want to have the LLM process, queue it up, and go back to what you were doing. Once you've finished reading the next article on HN about a 5 tps Xeon, your task will be complete. It's kind of like using a 3D printer: It doesn't matter if a print takes 10 hours, because when you come back in the morning it will be done.

Yes, with top-tier GPU farms you can hit hundreds of tokens per second. But if the old Xeon in the closet can get useful work done at 5 tokens per second, there are lots of people and lots of use cases where a free, unlimited 5 TPS stream is worth more than paying a dollars per day to get access to a 500 TPS source.

[RussianCow]

Once you've used a model that runs at hundreds of TPS, it's hard to go back. Everything completes so quickly that you can iterate without breaking out of flow state. My biggest gripe with slow (<50tps) LLMs is that I've lost all the mental context I built up by the time it's done, which makes it extremely difficult to explore or iterate on solutions.

[nmfisher]

Completely agree. Slow but smart models (Fable, Sol, GLM5.2 etc) are great, but they leave me with zero mental model of the code that's been written. Most of the time my mind wanders off and I go check social media or fire off a prompt for some other random project, it's a big productivity drain.

Working with models that are super fast, but slightly dumber (like mimo-v2.5-pro-ultraspeed) is amazing, I feel like I'm still the one that's actually making every decision.

[trollbridge]

Glad to see another UltraSpeed addict here. I really hope they keep it around.

[ngcc_hk]

In 1980s ibm has studied and said why sub-second response needed to maintain the mental flow. That time you send a whole screen unlike unix like character by character. This proves very true even when you deal with form processing. I think that we are dealing with the same issue here.

Keep your mental context in your brain is critical

[hx8]

I'd rather have slower and better output than worse and faster output.

[RussianCow]

It depends. For something high stakes or inherently complex, sure, you don't want to have to clean up the agent's mess afterwards. But for many tasks like building web UIs, the difference in output quality is going to be small enough that iteration speed will win over quality.

With a fast enough model, I can iterate on the UI of a given screen 4-5 times before Opus finishes its first attempt.

[allknowingfrog]

We clearly have different goals. I want an LLM to review my code, not the other way around.

[pmarreck]

I'm sure this exact topic has been argued hundreds of times already on HN, but I think I have a new "possibly agreeable to both sides" perspective on this after having lost man-years to retired corporate code aka "FAIAP, throwaway code"

Let LLMs write the corpo code, as it will be unlikely to still be running in 5-10 years. Frontier AI is already at the point where it writes fewer bugs per LOC than humans. By a lot.

Go ahead and do your bespoke coding on your side-project loves and core libraries... The stuff that will last, anyway.

But if you're working for a corpo and still doing bespoke... That's... not gonna last, I'm afraid. Well, either you remaining there, or that, as it were.

[allknowingfrog]

There's a whole spectrum of employment between faceless corporations and personal side projects. AI will replace humans because giant business believe they can do the same work, not because they will actually be able to.

The correctness of an application is limited by your ability to understand and describe what you need. We have a word for an application specification tool so detailed it eliminates all ambiguity. It's called a "programming language".

The mistakes are always in the transfer from human to machine. I still find a high-level programming language to be the best way to express my intent. Humans will make mistakes in the hand-off to AI just like they make mistakes in the hand-off to code, but at least code is deterministic.

[pmarreck]

> AI will replace humans because giant business believe they can do the same work, not because they will actually be able to.

This has been being claimed for at least 2 years now. Wouldn't we already be seeing disasters if this was the case? It's certainly been around long enough to cause some real damage. Instead there is a slow trickle of things that makes the news by people who didn't bother instituting a single fucking control. For example, I am completely immune to `rm -rf` style fuckups because I wrote rm-safe years ago and it is mapped to `rm` in any environment that the LLM will run in: https://github.com/pmarreck/rm_safe

In any event, it's not a replacer, it's an augmenter. Nothing will replace humans, because we are the stakeholders; it may shift them around, though.

> The correctness of an application is limited by your ability to understand and describe what you need.

If you write code in Lean 4 or Idris 2, you may not completely understand why it is or isn't correct, but their respective compilers will certainly prove it to you one way or the other.

We already are perfectly functional with incomplete understandings. An entire generation of web developers have had successful careers without having a single clue how compilers work or how machine code works. Now is there a type of problem that sometimes comes along that DOES require deep understanding? Of course, that's when you call in the heavy artillery. Or the advanced LLM. Or both. See my point?

[nolok]

It's still the same thing, you can ask it to do a full on report give explanation and details be thorough and then go do something else, another task a lunch break whatever and it will be done when you're back

[Aurornis]

> another task a lunch break whatever and it will be done when you're back

At 5 tokens per second and unknown prompt processing speed, you may need a very extra long lunch break depending on your codebase.

[allknowingfrog]

How do you maintain a flow state during a lunch break? I'm looping with Claude on a scale of minutes. While you're waiting, I'm iterating.

[vitally3643]

This is like comparing a hammer to a screwdriver and feeling smug because you can hammer nails faster than someone else can drive screws.

These are fundamentally different tools for entirely different applications. They only look similar to people who don't understand the tools or their purpose.

[fwip]

You don't "maintain flow." You eat lunch.

I swear, tech culture has gotten people wanting to work for the machines, rather than the other way round.

[Aurornis]

> Once you've finished reading the next article on HN about a 5 tps Xeon, your task will be complete.

If I spend 10 minutes reading an article, that would only generate 3000 tokens.

That’s not counting the prompt processing time.

We have very different expectations for LLMs if your tasks only take a couple thousand tokens and you’re happy waiting 10 minutes for it.

> Yes, with top-tier GPU farms you can hit hundreds of tokens per second

My 5090 gets hundreds of tokens per second with this model. No farm needed. I’d have to double check but I think even a $1000 Intel B70 might break 100 tokens per second.

> But if the old Xeon in the closet can get useful work done at 5 tokens per second, there are lots of people and lots of use cases where a free, unlimited 5 TPS stream is worth more than paying a dollars per day to get access to a 500 TPS source.

If that old Xeon pulls 200W from the wall and you pay national average electricity costs, it’s going to cost $0.90 per day to run it.

I would rather pay a dollar per day, get my answers 100X faster, and not have an old Xeon heating up my house.

[zoobab]

You could suspend it to ram, and only wake it up on request, it takes 2 seconds on my box.

[Aurornis]

It’s not a cost savings relative to paying API prices even if you’re suspending it.

This is an option if you must run local inference, you’re not sensitive to speed, and the budget is low.

It’s not going to be cheaper than paying API prices for the model though.

[dofm]

> It's kind of like using a 3D printer: It doesn't matter if a print takes 10 hours, because when you come back in the morning it will be done.

This is how I used to think about my 3D printer, but FWIW the way my actual thinking and planning works, print speed really matters. Not for the final print, but for iterative work and test parts, it is obvious that either having a fast printer helps. Having multiple slow printers also helps, but there are only so many areas of a design you can iterate on at once.

At the moment my own LLM use is experimental and iterative, and I definitely favour the faster MoE models for much of what I am doing, even if I might in principle prefer to get the final work done in the slower ones.

[adastra22]

We aren’t there yet. Not for frontier development work at least.

[varispeed]

Filament snaps at 1am and then you have to run print again. 10 hours turn into many days potentially.

I watch tokens to see if it goes in right direction. If model goes off the rails, then it is time to stop and adjust prompt.

[parineum]

> there are lots of people and lots of use cases where a free, unlimited 5 TPS stream is worth more than paying a dollars per day to get access to a 500 TPS source.

That Xeon is going to be devouring power.

[bitpush]

is there a good tool to manage these workloads? batch process a bunch, handle failures, retry things etc?

[bredren]

Except often queued agentic flows must be checked in on. Or to use the comparison, 3D printers are not immune to making spaghetti all night when something goes wrong. (I’m not a 3d printing expert so maybe that is solved now)

It is common for agents to just stop because overload or some API error hijinks.

Or you get a TUI question that is blocking.

In general you’re right though, staring at tokens from agentic is not time well spent.

Some of these I’ve built custom harness around in iterm2 though.

[bigyabai]

Yep, and we don't even know how long they spent on prefill. A typical 50-100k token session could take 10-20 minutes to prefill on a Mac.

Pending any new hardware or radically different LLM architectures, we're going to be waiting a lot longer than 2027 for 200B models on local hardware. SOC platforms like Apple Silicon are hamstrung by their obsession over raster performance, the hardware lacks the fundamental GPGPU hardware to be a replacement for real-world inference.

[aziis98]

For me, at least for agentic use, you need at least ~40tps. Less might be good only for tasks you could run in the background (like at night maybe).

Instead of coding agents like tool for local models I would like to see more "docker ps" like tools where you can queue up tasks that get processes incrementally (maybe when the pc i idle) and are specialized for doing retries and caching as much work as possible to work decently with slow local models.

[kevincox]

I think 10 t/s output is usable for many coding workflows if the input speed is much much higher (~100 t/s is a rough minimum line). The low output speed can really hurt for heavy reasoning output but it can still be used to get some jobs done as long as you don't expect highly interactive use.

[ClikeX]

The slower models seem fine for home lab usecases such as processing document transcriptions and tagging them, for example. I don’t need that to be live, it can just churn overnight.

[aaulia]

I tried Qwen MoE a while back. Using my 8GB RX470, somehow got 10+ token/sec, lot's of trial and error with llama.cpp config, and it's still slow to be used for my usecase. Even at 12 to 16 IMO it's slow. For chat, maybe it's enough, but for any other tasks, it's not viable IMO

[dwa3592]

i am working on making it faster but to me 7-9 tokens/sec feels very good. it was 0 tokens/sec a year ago.

[jckahn]

Ignore the haters. What you've done is incredible!

[Aurornis]

If you’re interested in these projects you should check out the project this was based on: https://github.com/JustVugg/colibri

It says so right in the readme. They’re not hiding anything.

[Wowfunhappy]

How fast does a human write code?

[sigmoid10]

It's fast if you want to automate things that run independently or overnight. It's slow if you want to iterate code together with it.

[Wowfunhappy]

If you were iterating on code with a human partner, how fast would they write?

Like, obviously faster is better and the entire point of computers is to do things faster, but I find it kind of surprising how many people consider 9 tokens per second—heck, even much less than that—to be unusably slow. It's still automating a class of task that virtually no one before 2022 was able to automate, and it's faster than basically any human can write code.

I guess the better question is, why would you use a 9 token per second system when you could use a much faster cloud model. Obviously, if you want speed, that's the way to go. But a lot of people seem to find the idea of sending their requests to a third party server untenable. If 9 tokens per second is the best you can do—I don't know, that seems usable and useful to me.

[mh-]

This isn't just measuring the speed of tokens that are written to a file (code). Your human partner is thinking a lot faster than 5 tok/s.

[sigmoid10]

I guess you haven't used Codex/Copilot much? All good LLMs now use reasoning. Especially for code it is almost a requirement to produce correct results. But that means even having them implement a comparatively simple function or algorithm can easily produce 3000 output tokens. Waiting 5+ minutes every time you iterate because you want a tiny tweak in the code is not a practical strategy for live development.

[imtringued]

You have to remember that the CoT thinking tokens take a while to produce. If your model produces 900 thinking tokens you're waiting almost two minutes for the first token.

The biggest argument in favor of running local models doesn't seem to be privacy at all, it's the fact that you can't run out of tokens even if it is a bit slow.

[notnullorvoid]

I mostly agree with the prediction though maybe a bit more pessimistic about the timeline. Also I'm not sure our current usage of parameter count would make sense in this scenario, such a feat would require compressing current parameters in a manner much different then something producing a bit count per parameter. A hypothetical example would be using a single seed parameter per layer which then passed into a noise function produces the functional weights for that layer, able to reduce per weight size to sub bit levels (256 bit seed, producing 16K weights).

[razster]

I sense the same. Running Ornith 35b with Pi and hitting 50+toks, and now that I learned Pi can do search and fetch, I have not needed to visit any of the large models as a search function. Soon we will be seeing new models drop this month and next that will change some of the landscape. Exciting times. P.s. Try Ornith, well worth it.

[cyanydeez]

setup deer-flow: https://github.com/bytedance/deer-flow

its pretty good for building any kind of research report using search engines and it manages context corrextly.

[Aurornis]

> I have a prediction. By the mid of 2027, we will have >200B MoE models running on basic consumer hardware.

This prediction alone isn’t useful at all without a bound on speed and maybe quantization.

You can already run >200B MoE models on basic consumer hardware by picking a low bpw quantization and then streaming the experts from SSD. There have been a lot of proof of concept demos, but nobody uses them because they’re so slow and the quality is so degraded.

If you’re saying that hardware will catch up by mid-2027, I disagree. The limitation is fast memory and that’s going to be expensive for a while. I have a 128GB unified memory machine that can technically run 200B MoE models with enough quantization, but it’s so slow that there’s no reason to do it. It’s going to be a few years before we have enough RAM and processing power in basic consumer hardware without spending as much as a used car to get it.

> This is a GPT4 level model running locally with a decent speed on a 16gb ram macbook air.

Sorry, but 9 tokens per second with a slow prompt processing speed is not decent for anything other than getting short chat responses.

You’re also not running the full GPT4 quality model. I’m very familiar with that model from some other work and the 4-bit quants are just not as good as all of those KL divergence plots would have you believe.

You also have very short context. It’s basically useless for anything more than short chats where you’re okay watching the output come back at reading speed, skipping the reasoning part (which is important for calling it GPT4 level quality), and waiting a long time for the first token.

Yes, it’s technically running, but not in a way that would be useful by normal LLM standards.

[dwa3592]

>>Yes, it’s technically running, but not in a way that would be useful by normal LLM standards.

What are the LLM standards?

Do you know how many people use perplexity? I know many people who are not software engineers or tech workers and have a LLM subscription for rewriting their stuff (non-native english speakers) in english. There are many use cases for running good models locally. Maybe not for you, but someone might find this beneficial.

[Aurornis]

I have a free perplexity account from some promotion. Not sure what comparison you’re trying to make because Perplexity’s whole thing is that it’s really fast. It launches the search with parallel agents and then even seems to render some of the output paragraphs with parallel sessions to get the results.

Doing the same thing at 7-9 tokens per second, concurrency of 1, would take ages for all of the tool calling and subsequent processing.

It wouldn’t compare in any meaningful way, because perplexity delivers instant results. That’s what I meant by modern standards of LLM usefulness.

[dwa3592]

Its really easy to argue against local models because when it comes to quality, you can argue using the tokens/sec. and when it comes to speed, you can argue using the parameter count. This is not compared to the frontier stuff but it is the frontier of last year that now runs on a local machine. It was impossible to do this last year.

[andrekandre]

the local models open source harnesses are really improving quite fast; just a few months ago i couldn't get any tool calling to work, and responses were very slow (thinking going on too long etc) but now with some newer models on my macbook air, tool calling works and depending on the model, it returns from thinking fairly swiftly....

[zozbot234]

> You can already run >200B MoE models on basic consumer hardware by picking a low bpw quantization and then streaming the experts from SSD. There have been a lot of proof of concept demos, but nobody uses them because they’re so slow and the quality is so degraded.

On the contrary, that's ultimately an excellent baseline capability for most casual LLM users. Especially if you add the ability to fire off multiple requests over time and work on them concurrently (at least for not-very-long contexts), which is quite natural if one can implement some sort of continuous batching.

Even quantizing the model (with the ensuing loss in quality) is not an absolute requirement, quite unlike e.g. ensuring that a local model can run fully from VRAM on a typical consumer dGPU.

In general, ISTM that people overemphasize real-time or near real-time response which is a rather terrible fit if your goal is efficient LLM inference of near-SOTA models on typical local hardware.

[imtringued]

>If you’re saying that hardware will catch up by mid-2027, I disagree.

If DDR6 comes out that will actually double the effective memory bandwidth on next generation computers. Most computers will reach a good fraction of a Strix Halo system's memory bandwidth and the next generation of Strix Halo will reach Macbook Pro levels of memory bandwidth. Of course there is no guarantee this will happen by 2027 but DDR6 will probably exist in some form by the end of 2027.

[VladVladikoff]

Isn’t 9 tok/s unbearable? I frequently see Claude sessions hit 1Mtok in less than a day. It seems that 9tok/s would be really slow for actual work.

[driverdan]

You can't use it real time but you can have it run in the background and come back to it.

[mseri]

Nice!

I had done the exact same with gemma4 26b, both for my Intel laptop and for my M1 with 8Gb RAM (with also q4 and turboquant). I don’t use it much since there are dumber but way faster models to run, but I should clean up the code and make it available

[mseri]

Ok, done: https://github.com/mseri/zunzuncito My main focus is systems with very low ram. On my M1 with 8Gb or RAM llama.cpp cannot run gemma4 26b, but this implementation works fine with 5-6 tok/s.

I also replaced my crap implementation of the oai server following yours, I hope it's all right. I did add a mention to your project and this fact in the README. It is a lot more barebones than what you have, but I have to admit that it works really nicely for me

[dwa3592]

glad to know. thanks :)

[felineflock]

That is awesome!

I am curious about the decision to not use GPU since this is for Apple Silicon.

Wouldn't the GPU potentially accelerate the DeltaNet/attention layers and matrix multiplication in general?

[dwa3592]

It will, but the process at this point is SSD bound rather than compute bound. On a bigger machine, Apple silicon must help but I don't have a bigger machine. I can think about this more and will make changes if that helps.

[damienmeur]

Just wake me u when I can run fable or gpt 5.6 for free my raspberry-pi 4 at 2000 tokens/s

[haute_cuisine]

How are the thermals? I noticed that running any serious workload locally heats system fast.

[dwa3592]

i have been optimizing for that. for now samosa is capped at using half of the avaiable cores and switching between them, which keeps the system 'less hot' as it would have been. i will also release better thermal control in the next release. at this point its basically sacrificing about 20% of the speed to keep the hardware less stressed (and hot).

[Capricorn2481]

I tried Qwen3.6-35B-A3B, but it couldn't generate a 50-100 line Clojure file without having broken parens mismatches. I know Clojure isn't super popular, but the syntax is pretty simple and the frontier models do fine with it.

[segmondy]

try q8, check your parameters. qwen3.6-35b-a3b should definitely be able to do so with no issues at all.

[Azantys]

You are comparing a 35B models to a 635B+ frontier model, of course thats not even close

[Capricorn2481]

I'm not lamenting that they aren't close, I'm saying Qwen will frequently output code that isn't even syntactically correct, even when the syntax is simple. Which makes it unusable for coding.

[IsTom]

To be fair, they don't have the text editor highlighting all the matching parens. I'd be lost too.

[Capricorn2481]

Yeah prediction models and many parentheses are probably not a good combination, but we're not talking about anything exceptionally complicated here. I have had syntax issues in Python as well.

[Azantys]

It really depends on the language, popular languages work pretty good

[walrus01]

what quantization? what temperature?

[minkeymaniac]

I just downloaded llama. ran this llama-cli -hf ggml-org/gemma-3-1b-it-GGUF

I am getting [ Prompt: 91.3 t/s | Generation: 171.8 t/s ]

This is on a GPU (RTX 4060)

Is this decent?

[typon]

By early 2028, major players like Intel, AMD, QC will ship accelerators in consumer laptops capable of running ~1T MoE models at ~100 tok/s

[veber-alex]

Literally the only way this is going to happen is if aliens come to earth and gift us some amazing technology.

[typon]

Yes that's called Mythos 2 or GPT 6

[kennywinker]

Unless there are major improvements to how much hardware it takes to run a 1T model, this is deeply unrealistic. First because why release hardware that puts your biggest customers (data centers) out of business. Second because as I understand it the data centers have bought up all the high end chip production capacity for at least the next year and unless the bubble pops that'll continue for a while.

[iwontberude]

Because for the company that will actually do it, their biggest customers aren’t data centers they are iPhone owners.

[kennywinker]

First off the math doesn’t math. Datacenters are willing to pay $50k for a single high end GPU. If you have unlimited capacity, yeah sell millions for $100 a pop or $10 a pop or whatever the bom cost of a phone GPU would be - but if you have limited capacity, you’re gonna sell all of that to the customer who is willing to pay the most PER UNIT.

Second off, this doesn’t work from a power consumption standpoint. When I run qwen3.6-35b, a far smaller model than op is suggesting, power usage spikes to 150-200W during inference. To fit a 1T model in the palm of my hand, the amount of processing required doesn’t fit the amount of power available.

Now I’m not saying this will never happen - there are some great leads, e.g. burning models directly on to a chip - but op’s scenario is definitely not happening in two years. Maybe 5, a lot more likely 10, unless of course local ai is made illegal

[typon]

There is ton of room for improvement "down there".

* Software inference optimizations

* Heavy quantization

* Chips with hardcoded transformer architecture

* Much cheaper HBM

* Much sparser models - 1T total with ~1-10B active params e.g.

* Not to mention - 2 years of today's frontier models writing RTL and kernels at superhuman levels.

[kennywinker]

> * Software inference optimizations

Absolutely. I'd be surprised if they couldn't 2x performance in the next year. Still doesn't make a 1T model fit on your phone.

> * Heavy quantization

I think this is a dead end if you're trying to fit a 1T model into a phone. Makes much more sense to train a model that's designed to be small, than train a model that's smart and then quantize it into stupidity.

> * Chips with hardcoded transformer architecture

Totally, this will probably work great. Now good luck booking fab time any time in the next 2 years.

> * Much cheaper HBM

Totally, this will probably work great. Now good luck booking fab time any time in the next two years.

> * Much sparser models - 1T total with ~1-10B active params e.g.

Fewer active params helps with the speed of token generation, but if the whole model doesn't fit into ram it doesn't solve the issue of having to constantly stream portions of the model from disk to ram.

> * Not to mention - 2 years of today's frontier models writing RTL and kernels at superhuman levels.

IMO this is a delusional myth-making idea being sold to us by ai companies. Machines that generate output based on statistical averages won't generate genuinely new ideas. They can help us try out ideas faster, but they're simply not capable of the kind of creativity and understanding required to push a field forward, except incrementally.

[iwontberude]

You are assuming people need the models they use today. The reality is much much smaller models will suffice (i.e. dont use god models for dog work)

[kennywinker]

I’m not assuming that at all, I’m responding to someone suggesting we’ll be able to run 1T models on phones in 2-ish years.

I absolutely agree that models are going to advance on to “edge” hardware over the next few years by becoming small + specialized.

[andrekandre]

  > Datacenters are willing to pay $50k for a single high end GPU.

its true for now, because capital is flowing like a torrent, but how long will that last if returns start to be expected (aka the bubble pops)?

[kennywinker]

Even if the bubble pops and anthropic and openai et al implode - genie doesn’t go back in the bottle. The usefulness of LLMs for coding is proven, and a chip in a datacenter running 24/7 is always going to be more valuable than in a personal device running occasionally.

That doesn’t change until production capacity exceeds the datacenter demand. When that happens, they’ll start selling them down the market until it eventually reaches phones and toasters and whatever. But not in two years.

[javier2]

How do I install without curl-piping to bash?

[TYPE_FASTER]

Downloading now just 'cause the repo name

[rvz]

> I am running Qwen3.6-35B-A3B locally on my 16GB mac with 7-9 tokens/second.

That is no where near decent at all.

[dwa3592]

it's a 16GB machine. i am proud of this machine so far.

[embedding-shape]

> I am running Qwen3.6-35B-A3B locally on my 16GB mac with 7-9 tokens/second. Link - https://github.com/deepanwadhwa/samosa-chat

It's quite telling you didn't use Qwen3.6-35B-A3B locally to build that, seems there was another collaborator ;)

Show something you've built with the model+tooling instead, truly dogfood it. I'm sure you'll discover things along the way too!

[smith7018]

Nothing says they're using Qwen for local development. They could be using it to for conversations, knowledge, or "creative writing."

[embedding-shape]

> Nothing says they're using Qwen for local development

I know! That's my point! You're a poor salesman of a coding environment/tool if you don't even use it yourself for coding...

[dwa3592]

>>It's quite telling you didn't use Qwen3.6-35B-A3B locally to build that

that would have run into a race condition unfortunately ;)

but there is a sample landing page + a python function on the repo which shows what the model produced. my goal is to integrate the local model in my workflow so that claude/OAI can call this model for basic stuff.

[embedding-shape]

> that would have run into a race condition unfortunately ;)

Not really, you start small, bootstrap as soon as you can, and off you go. Requires a good model though ;)