I'm surprised neither OpenAI nor Anthropic made this move first. The Chinese open weight models are pulling ahead and commoditizing their value proposition.
Baking models onto silicon would've been the next logical move to get a moat.
Google is already doing this and has an experimental project on top of already having TPUs and cramming their quantized flash onto individual TPUs for inference.
This is neat but IMO a little crazy.
Something I personally haven’t seen much of, in all the discussions of model benchmarks and AI breakthroughs, is a distinction between “peak performance” and “reliable performance”. The “peak performance” of frontier models is very high: they’re solving open math problems, analyzing large codebases, etc. But my subjective impression is that “reliable performance” is mid at best: out of 100 random questions I might think to ask, it’s likely to say something wrong or stupid a handful of times at least.
I think there’s inherent tension between the two: the more a model reaches or outright hallucinates, the more likely it is to come up with tricky, subtle solutions to problems (I think people are somewhat like this too: Terry Tao’s brother is nonverbal, Jim Watson’s son has severe schizophrenia, etc). But then the less likely it is to generate a sensible email reply.
I use models all the time for coding, but I would not let one take over my daily correspondence. If the idea here is to run frontier models at high speed in data centers, that could be useful (the speed would be cool), but I’d be surprised if the cost of that hardware churn is worth it to frontier labs. But if the idea is to turn this into a chip that goes in your phone as some kind of routine, low-power inference thing…taking something too kooky to be relied on and baking it into your phone’s hardware like that doesn’t make sense to me.
The demo: https://chatjimmy.ai/
I know it's a relatively tiny model, but damn, is that thing fast.
It also mostly passes the "schlong" test
I read the paste, it got the etymology wrong, no? Schlong comes from shlang (snake), not shlemp (is this even a word? I don't speak Yiddish but couldn't find it on Google).
Oxford also claim that its first recorded use was from the 60s, not the 20s; https://www.oed.com/dictionary/schlong_n?tl=true
It did get it wrong but it also got a lot farther than much more recent, but worse models like 6.7GB on disk size ternary bonsai. It at least knows it's from Yiddish. The "schlemp" appears to be a total hallucination or it's confusing it with schlep, which is not related to schlong. One of the reasons why I said it "mostly" passes the test. Something much larger on the size of qwen 3.5 122B, deepseek v4 flash or similar that runs in 120GB to 190GB of RAM in my experience will answer perfectly unless it has been ruined by something like Q2 quantization.
I didn't realize there was a SchlongBench™ (but of course there is). What's it test? (asking seriously)
There isn't SchlongBench(TM) yet, it's a specific question I've been asking of differently sized models as a randomly chosen gauge of how much less commonly used knowledge is perma-baked into it. In this case a question about a specific yiddish origin slang term. Small/bad models don't know it's from middle high german or Yiddish and get its origin and meaning totally wrong (or it runs into model censorship related to slang related to the male anatomy).
It's also a question I have found will cause models that don't know what it is to go off quickly in a direction of hallucination trying to explain it, so the hallucination is evident very quickly starting from the first ever prompt issued with 0 context fill. Example: I had a model write four detailed supposedly-accurate sounding, grammatically correct paragraphs saying its origin is from AAVE (African American Vernacular English), which it most certainly is not
You could do the same by picking any topic that is very rarely discussed in conversation, some esoteric and narrow piece of knowledge and asking the model about it.
Oh ya, this is like the approach from the Incompressible Knowledge Probes [0] paper - smart!
[0] Incompressible Knowledge Probes: Estimating Black-Box LLM Parameter Counts via Factual Capacity [https://arxiv.org/abs/2604.24827]
I freakin' love this demo. It feels magical.
I had the same reaction but then I showed it to my partner. She completely didn't get it, in her words "how can it be thinking of a good answer when it's that quick?"
I tried to explain but I fear were probably going to be adding artificial sleeps to these things to convince the masses it's doing something clever.
I asked it some old hardware command line questions I'd recently asked Gemini, it hallucinated parts of the answer.
The characters in the 3-act Shakespearean play had very little depth, many of the names were similar, and they were not very smart, but the simple plot was cohesive.
to be fair, the model used for Chat Jimmy is not very smart, but the world where it is smart is very interesting.
It’s going to be really crazy when the bottle neck for agents is the speed of the tool calls rather than the speed of inference. Imagine an agent interacting with the terminal near instantly…
Wow, feels like Google web search in 1999.
If you still want the experience, go and browse McMaster Carr. Wizards designed that website.
OMFG this thing is fast.
its fast but try to get it to give you pi to 50 decimal places. it didnt go well for me.
I think the same exact model running on CPU-only and RAM, or a small GPU, would do about the same? It's quite an old model now and small, you could throw a GGUF into llama-server or something for a side by side comparison.
https://huggingface.co/meta-llama/Llama-3.1-8B
As I remember just about any english language model from mid 2024 and earlier didn't even do well if you asked it to count sequentially from 0 to 100, nevermind calculating stuff.
I understand the appeal due to the speed
It doesn’t believe it’s running on that chip, it’s arguing with me
AIs don't intrinsically know anything about themselves so they often give wrong answers to such questions. This can be fixed by putting info in the system prompt but they may consider it a waste of tokens since most usage doesn't benefit from that information.
It's running a very small, non-reasoning model at the moment. But more generally, almost all LLMs argue on the hardware/model they are/are on.
What would tokens/sec performance look like for a reasoning model? An order of magnitude slower?
Reasoning models are the same speed. They’re just post trained with RL to do CoT inside tags like <thinking></thinking> before a tag like <response></response>
There’s no difference in the inference implementation, parameter count, or speed.
Which model? Or how many active parameters?
So this demo is around 90 times faster than typical speeds for the same model at openrouter, and around 30 times faster than the absolute fastest option available (Groq).
Field reprogrammable, it's an FPGA on steroids. Field upgradable.
Burnt in, it needs a zif socket and easy access in every car, aircraft, a pull out slot in a phone, or it's new era planned obselescence.
Can that be done when the whole idea is to store a multiplier into a handful of transistors?
Given the fast churn of the models, how does it work out?
Won’t the silicon etched model already be 1 or more versions behind by the time the silicon comes out.
Though if it’s cheap enough, there certainly can be a market for cheaper model inferences.
I find speed alone would be a game changer for current models. I hardly find any task anymore that the current frontier models can't do with max reasoning after several rounds of feedback (provided sufficient instruction and the right harness). But waiting an hour or more for reasoning to finish is getting really cumbersome. If they could do the same in seconds (and for cheap of course), I'm pretty sure we'd pretty soon see major software companies pop up that are run by a single human.
This is probably a win-win. The team gets paid, and we get greater assurance that their best ideas and architectures -- which are truly impressive -- are going to see the light of day in actual products.
They were too small for this to be a meaningfully sized purchase for AMD, there's real risk they get sucked into a team that ultimately delivers sqat, not to mention the chances of anything being delivered in an even remotely consumer-priced bracket are definitely out the window
It obviously won’t be continuous delivery but could make sense if the lifecycle of a model (train, deploy, iterate (meaningfully) is about 1-2 years. In that case it fits nicely in the “this year’s model” already established with cars, phones, etc.
Can anyone comment on the economics and likely turnaround times of this process, when it’s more mature?
Would it be realistic for a frontier lab to deploy this or would the turnaround time mean the model is always too out of date?
Assuming the weights and architecture are eventually stable, how much cheaper would this end up being?
There are always uses for outdated models.
Claude Code is still using haiku 4.5 from ages ago for explore subagents for instance. Not to mention production uses like customer service that only need to be "good enough"
Customer service has really degraded huh. 4 years ago they expected opus performance out of human call center agents
I guess losing some customers due to poor customer service is ok if the price of customer service is right.
Just looked this up, no longer true. Explore subagents inherit whatever model the parent is. And you can of course make other subagent configs.
That's solely so that you burn more money. It's totally unnecessary to assume the parent model. Sure, it could be upgraded from Haiku if there was a solid reason to, but...
I mean, if you could get Opus or even Sonnet 4.5 at 1000+ tok/s exploring the codebase, they would probably change that setting back.
But either way, I think GP's overall sentiment of "delegating intelligence-saturated tasks to an outdated but fast subagent" makes a lot of sense.
2 to 3 months optimistically assuming everything goes smoothly and is fully automated.
6 months or even a year if something goes wrong in the fabrication process and you need to update things.
If they do more standard asic design, it could be a lot longer as the design needs to be validated on an FPGA cluster, which would necessarily need to be very big for something like a LLM. Easily up to 2 years.
There's a reason chatjimmy isn't demonstrating newer models and why they only show of an 8B model.
I mean even if it take a few months, it'll still be out of date. But there was a hypothetical when it came up in Feb, would you want Qwen 3.5 at like 10k tokens per second.
At the time people were no doubt saying yes but now 3.8 is out, is that still desirable?
There's soooo much stuff that such a model is still capable of doing in the pursuit of getting a better overall answer. Imagine a powerful research agent that blasts out dozens of the small, cheap models to fetch and summarize one page each. Then the beefy researcher model performs the final analysis.
AMD could have saved their money and used their own hardware! I've got a language model doing 60k tok/s on AMD hardware already, a Xilinx Kria K26 SOM, with the weights baked into URAM/BRAM with zero DRAM in the token loop. Same thesis as Taalas: single-stream decode is bandwidth bound, so stop fetching weights from far away.
Caveats stacked high, obviously. It's 3.16M parameters (tinystories, and I also have a kevin-speak lemmatised version), the tokens are characters, and the 60k record is 16 streams that each remember exactly one token of context, so it's blisteringly fast at saying nothing. The honest build with full context and KV caching still does ~19k tok/s on one stream though.
I keep messing with the blogpost with the live demo, but I'm planning on flipping it to live in the next day or two
[delayed]
Well, technically it is their hardware now...
And their team, if they treat them well.
Really hoped to see their hw out in the wild one day
Well so much for that dream.
Guess we can look forward to picking these up ex-enterprise on ebay for under $5k a pop in a decade or two
Honestly, this is starting to make more and more sense. SOTA models are starting to converge to certain architecture and capabilities. I wouldn’t be surprised we end up with a base model ASIC + “fine tune” card where it’s a physical LoRA style adapter.
Imagine a multi-modal model with 1000's of tokens per second. Realtime inference for a host of applications. This is a BIG deal and will change the landscape in unfathomable ways.
The https://chatjimmy.ai demo was impressive.
Once models settle down this makes sense. Imagine a cartridge with a physical model on it. You purchase a cartridge and stick it in your computer/phone/server. Want to upgrade? By a new 'cartridge'.
This should bring inference cost down dramatically, I wonder how OpenAI/Anthropic feel about that.
i'm looking forward to Qwen3.8 27B launch to see how much models have peaked at a given size.
it might already be time to start burning the best small models onto hardware since it's possible they can't get much better at many tasks like knowledge recall due to the inherent information density limits for models at a given size.
very interesting idea. i didnt think of that. i was just assuming youd have an additional one of these in your phone for actual lightning fast local inference
> Imagine a cartridge with a physical model on it.
I can finally have my own Dixie flatline. Cool.
> Dixie Flatline
In case some did not know: also the movie (or TV series?) is finally happening.
# Neuromancer - Official Teaser https://news.ycombinator.com/item?id=49055037
Then we can have machine psychologists pull cards when they run amok.
we have not converged at all, if you look at how different the chinese models in terms of architecture you can guess that the labs are experimenting a lot as well. we are seeing all different types of hybrid architectures, different attention methods and so on. Of course on a high level its still a transformer but if you take a proper look we are seeing more divergence then a convergence.
Wouldn't this mean someone with sufficient hardware could lift the SOTA model weights off the chip? Or are you saying that these chips would only be used internally by these companies and not sold to the public?
I wouldn't expect companies not sharing their weights today to be any more likely to share them if they're on hardware, this doesn't sufficiently hide weights from a local user.
The weights are very unlikely to be on the chip itself. That wouldn't work for SOTA models that are terabyte scale, even quantized. This is probably an accelerator for specific kernels in the model, but the weights are likely loaded from memory. The chip may have SRAM to store some of the weights temporarily during inference.
At least in the case of Taalas the weights are physically encoded directly on the chip.
It’s composed of 4-bit multiplier cells that compute all 16 possible results in parallel. The top metal wiring layer physically selects the one that corresponds to a multiplication with that cell’s constant weight, and route it to the next layer.
I don’t get why this is an issue? You can run Claude/OpenAI SOTA models through Amazon bedrock. These weights have to live somewhere to run on Bedrock.
One idea would be to use an open model.
The technical aspects of SOTA models are not publicly documented. How do you know if something is converging?
SOTA American models are not. SOTA Chinese models are. From a physics aspect, closed source models cannot be too far from open source ones in terms of size. There’s only so much you can squeeze out a B100 style cluster even with fancy Dflash style diffusion model for the speculative model.
If we had deepseek v4 flash 0731 etched on a chip it would be more than capable enough and fast enough for so many people's needs, even hardcore engineer.
Will be capable and fast enough for 2-3 weeks until new sota drops
I think it’s tongue in cheek. When I first got access to Sonnet 4.5 I remember thinking to myself “y’know if they never got any better and I just had access to this forever then that would be pretty okay”. Turns out my expectations have changed since then and I would like a higher baseline now.
if capability is a commodity then the differentiator becomes taste.
Because new stuff instantly makes anything prior bad and incapable and garbage of course! Did you forget the hype-machine speaking notes??? /s
if they were still exponentially increasing, they wouldn't be preparing for an IPO. IPO is where companies go to die and founders escape.
Having a base model ASIC as a physical piece of hardware makes me think of the early days of microcomputer desktop stuff where having a socketed ROM or PROM was a key piece of hardware, and people actually knew/cared what ROM was on their system's motherboard.
Imagine if like instead of having a specific Mac Plus ROM, you had a thing that looks like a fat ASIC that can hold models sitting on a slotted daughtercard directly next to the CPU and RAM.
I don't think there'll be a fine tune card; you'll have the base model vintage whatever year, and then your GPU will do whatever LoRA layers you want it to do; the LoRA will wrangle older dated models into the current of whatever your looking at.
But yeah, for things like programming, if it can do linux and python and some go and sql and javascript, larger domains can be threaded with LORA
Toronto Canada startup btw.
Works well, I remember driving by the ATI building as a kid.
Seems to be somehow some kind of offshoot from or connected to Tenstorrent, which is just down the road. Founder looks like he was/is maybe at Tenstorrent and previously associated with Keller?
Always fantasize about applying at Tenstorrent, but wrong side of Toronto. 2 hour commute.
While this design is self-limiting I think its a good approach. It doesn't take an entirely new architecture or infinite memory to produce significant performance improvement.
> At 20 billion parameters per chip, you’d need just 50 accelerators to support a trillion-parameter model
I don't see any evidence that this is possible. From my understanding, the whole model needs to be on a single chip. Which rules out any popular frontier models with several trillions of parameters. Even smaller sub-frontier models have hundreds of millions of parameters, so these would be ruled out as well.
With web search and tool call a decent current generation model at the speed of the chatjimmy could do a lot. People saying it would be out of date are missing the point. It’s not going to make much sense for frontier companies that’s chasing the SOTA. But for a lot of business use cases if someone can put GLM 5.2 and sell it as a box, it would make so much sense.
My partner has been asking for a “completely private” model for doing research and shifting through volumes of data that can’t leave the office and $$$ for the current hardware makes no sense. It would be an easy sell if someone walks in with a black box that contains “ChatGPT”.
100% agree - you don't need the most up-to-date model to have something that's useful in agentic contexts. They could even produce chips with weights that make all the decision making/logical reasoning and have it delegate to other specialized agents. If it becomes cheap enough to print a run of custom chips, releasing a batch for each major advancement does not seem unreasonable for SOTA companies.
There are so many use cases for supremely fast offline models. The first thing that comes to my mind is for real-time video processing or other non-textual content in real time.
I feel like NAND process tech could become useful at solving some of these problems. A GPU where you can update the weights a few thousand times may be sufficient.
The basis of Taalas is "compute in memory" electronics - past Von Neumann's separation of processor and memory.
You need to be able to add|mul where the data (the weights) are stored.
NAND hasn't been scaling great lately. It seems like PCM or MRAM would both be better fits.
FPGA model storage?
It must be a “super model”. What will be if new model released? New chips?
Didn't even give them a chance to launch the hardware.
so qwen3.x-27b on hardware? or better deepseek-v4-flash on hardware .
I wrote them an email asking for PrismML Bonsai 27b Ternary which is like 6b or something crazy small and would be a lot easier for them to do initially.
They were specializing their forthcoming system on 4-bit FP - which I understand is a structural decision.
Bonsai Ternary (1.7bits/weight) is a compromise, compromise that has to make sense in the context - efficient when translated into transistors.
Token quantity will have a quality all its own.
Imagine the size of chip needed to 'etch' something like Qwen 3.6 27B in size.
Interesting thought, because it's a yield question. How tolerant are models today to a few broken weights.
If tolerant, they could churn out many cheaper chips, some perhaps with slight abnormal tendencies ;)
> How tolerant are models today to a few broken weights.
Extremely! You can remove entire layers and the model will still work just fine, with barely perceptible capability losses.
I've cut/bypassed ~15% of total parameters out of Gemma 4 31B on a pod once. Still got perfectly coherent responses out of it. Certain layers are a lot more important than others, particularly early and late ones; but it's honestly astonishing how much can be cut out from the middle without destroying the model's coherence.
I didn't run any meaningful benchmarks, so I have no idea what the capability loss looks like exactly. But "produce coherent and sensible English in response to a wide variety of prompts" was definitely not among the things the model unlearned.
Brings to mind the scene in '2001' where Bowman is pulling out individual pieces of hardware that represent the mind of HAL, and it becomes increasingly incoherent as more physical hardware is detached.
I wonder if you had a few percent of problems in the yield, if it would be functionally equivalent to the difference between a unsloth-published Q6 standard size GGUF vs. the nearly perfect precision of an unsloth Q8-K-XL. Or more like Q4 vs Q8 where a lot is lost.
Not too dissimilar to the first HC1 (6nm 815mm² 53B Transistors embedding an 8b LLM):
> Our second model, still based on Taalas’ first-generation silicon platform (HC1), will be a mid-sized reasoning LLM
If someone has that sort of knowledge; how big a chip would be required? Is it possible?
Well, given the data above, roughly a 220b transistors chip for the HC1 tech.
Personally I think Apple should have acquired them. if you could burn a gemma4 class model into an iphone and actually get extremely low latency and low battery usage it would feel like the future IMO. even if it means you wont get frontier intelligence, there might actually be incentive to buy a new mobile device every year again.
The Taalas chips are not physically small. And part of their secret (if you look at the design) is just locating a bunch of memory soldered on the edges ( I belive higher amounts of SRAM ? )
I don't think this works out from a cost/silicon perspective. Small models already run pretty well in software (since the weights fit in cache) and big models require silicon area proportional to the size of weights. On a mobile device putting a chip like this is competing directly in BOM and power against a whole lot more l3 cache, and the l3 cache makes everything faster
The weights might fit in cache, if you're using a small model. If you wanted to have a 20B+ parameter model, that's just going in RAM. You could put more RAM in the device and pay the perf cost or have a dedicated chip. Most devices already have a dedicated chip, this just changes which silicon you're spending the money on.
My question is what changes about LLM use cases when you’re getting 1000 tok/s? Models in silicon might dramatically change how we think about them.
That's actually a really good point... There's currently zero incentive to buying more hardware, and that's one very good reason do have a new one.
Apple is somewhere between fashion company and second rate tech company.
They could have 9 year old AI and still post profits.
Not sure if it's my pixel or android, but I made a randos jaw drop with what the crappy AI on android can do.
When are we getting android OpenClaw?
From what I remember, these chips are not mobile size yet
A small model would be. I think that’s more the point. It’s definitely not SOTA but it’s fast and energy efficient and local.
Is that analogue or are they baking floating points into the silicon?
Nope, a small model would be larger than the whole iPhone SoC.
Considering the rate of model development and rail hopping, seems like baking models into silicon is speed-running obsolescence.
If you’re only running models for frontier capabilities, yeah. For tasks where current models are smart enough, running them 100x faster is the most impactful improvement you can make. Consider all the things you could use a model for, but don’t, because the latency is just a bit too high.
Compute the cost of producing n of them devices, imagine a fair price based on that, and see if that local, blazing fast card* can be an asset that could be replaced periodically.
*(It's local: private files managing firm oriented. It's blazing fast: it can be placed into recursive, intensive local workflows.)
"seems like baking models into silicon is speed-running obsolescence"
Now maybe. When models are flying passenger aircraft, other prerogatives will assert themselves. When a 50TB ROM means you can impulse purchase a ChatGPT 6.3 xhigh that runs on batteries, yet more use cases will be apparent.
Well, 50TB ROM Taalas HC1 style would be apparently a 400000b transistor system through a chip sized 2.5 meters on the side... :)
Which is exactly what companies and shareholders want to increase sales.
I could see this making sense when model development start to settle down ... it's going to settle down, right? ...
Not sure. You can fix the transistors but leave the connections between them open for flexibility, so you only need to change the manufacturing process for the upper masks for every new model.
Surely that added flexibility negatively impacts the density/parameter count of the model you could etch?
Or do a hybrid
Look at it the other way: compared to the cost of training a model, the cost of making a custom ASIC is trivial.
Apparently Anthropic is moving that way: https://arstechnica.com/ai/2026/08/anthropic-confirms-plans-...
Isn’t that kind of useless for the stock? It sounds complicated, unlike having number of CPUs go up.
It’s like talking about anything else than Megapixels when everyone was convinced that megapixels must go up in certain periods of the smartphone boom.
ASICs is what took over Bitcoin mining, cheaper in all ways, and lasts longer than Nvidia GPUs for inference.
I’m surprised Nvidia hasn’t partnered to make a Claude chip yet. It’s a win/win you can license them out, sell them when they become obsolete, etc.
A model can't be updated, and a chip that is only relevant for 6 months at max?
Depends what you mean by relevant. If you use AI primarily as a search/knowledge engine, it makes no sense. If it's your capable assistant that has a lot of general knowledge, can do tool calls, and has a big context window, very doable.
Indeed, for some kinds of applications involving secure/legal data etc. I can see the consistency of silicon winning out, because it combines performance with immutability and guardrails in hardware. Some chips have write-once PROMs to store password hashes and similar, you could do the same thing with prompt hashing to absolutely force or forbid certain behaviors. A model that can't be updated is also a model that can't be hacked.
People already buy new phones every year, this just creates even more reason to do so
Because Openai and anthropic are not hardware companies. They outsource that to Broadcom and AWS' Annapurna labs.
OpenAI and Anthropic are both designing ASICs.
It's a terrible moat. You etch the silicon then nobody wants to run it in 6 months because models have advanced that much further.
This is only true for people who are solely focused on performance. There is absolutely a market for acceptable performance combined with predictability.
Not so if it's embedded in something smart enough for its intended purpose.
Think vision, spatial reasoning, speech synthesis, even some speech analysis. Think self-driving cars (and drones) that need 10x less power for the brain, and can think at 10x situation per second.
If a model is good enough today, it's still gonna be good enough in a year. Except you'll be able to serve it 1/100 of the price. Or 100x the speed.
It googles models suck