Too bad you probably need something like an 8x B300 node to run this. Black Friday is coming soon though, but I fear this open intelligence will be pacing anyone's liquidity frontier so I'll probably keep trailing the frontier for a while with nvidia/qwen3.8-27B-NVFP4 on dual RTX PRO 4000 in TP=2. I can almost fit 450k context in there.
Models this size aren't targeting hobbyists running models in their bedrooms on gaming/surplus hardware bought with disposable income.
This model is for researchers who already have racks of equipment, inference providers who sell access to models, and for businesses who must (or choose to) use on-prem models. All of which have to pay to use the model if their revenue rises beyond certain thresholds: https://huggingface.co/moonshotai/Kimi-K3.1-Pro/blob/main/LI...
Considering the likelihood of Moonshot being able to collect a judgement for violating these terms in an American court is basically zero, and the relative ease of finding the weights without the licensing agreement, "have to pay" seems very generous.
It baffles me that Western countries don’t seem to apply the principle of reciprocity in matters like this.
If Western companies can’t realistically win a lawsuit against the Chinese government in China, why should Chinese companies be able to do the reverse?
Consider the second order effects, there's a reason Western systems emphasise equality before the law. Of course most real-world implementations fall far short, but the system makes equality before the law a goal for a very good reason.
Consider the alternative option, we weaken equality before the law when dealing with non-Western entities who are bound by legal systems we consider on strategic grounds to produce inferior justice. Not only would this by definition be a weakening of our standards of justice, it creates an enormous moral hazard. Common law systems like the UK and US rely strongly on precedent, imagine the kind of unjust precedents which might be set if we do not apply equality before the law strictly. On top of that, having weakened standards of justice to frustrate foreign opponents, what is going to stop the political apparatus from weakening standards of justice again to frustrate domestic groups it doesn't like?
Even if it leads to disadvantage on the global stage, in my opinion the risks associated with weakening equality before the law far outweigh the benefits. A legal system with strong integrity is an asset, and it shouldn't be spent cheaply on short-term geopolitical gain.
> Consider the alternative option, we weaken equality before the law when dealing with non-Western entities who are bound by legal systems we consider on strategic grounds to produce inferior justice.
Maybe the alternative is equality-before-the-law, but the judgement amount won by the Chinese company gets put in escrow until there's legal reciprocity, to create pressure for it.
Holding payments in escrow has the problem of making the administration of justice contingent on the behaviour of a foreign government, which brings us back to where we started with equality before the law. How can we be said to have a higher standard of justice if we explicitly carry out the same kind of instrumental justice we're accusing others of?
Personally I think building trustworthy systems for international arbitration is the method most likely to succeed, if the aim is to come to a common standard of justice for commercial matters then courts on either side of the equation will always be considered suspect by the other side. Both credible international institutions and neutral third countries are a good way of doing this. Appeals to the virtues of different systems aren't going to work here, instead we have to create material incentives that favour a high standard of justice. A reduction in these kind of legal difficulties would improve trust and efficiency.
Obviously this approach has its limits but that's when things like diplomatic pressure, economic protectionism, and in the worst cases economic sanctions come into play. This is all the realm of the executive and legislature though, not the judiciary. The principle is not mere idealism, if we let day-to-day international relations contaminate the judicial process we are committing exactly the same sin that we're accusing China of. By keeping the instrumental aspects of foreign policy isolated in the executive and legislature, this problem is avoided in the interests of justice. It also helps avoid the perception of hypocrisy which is always a diplomatic advantage.
because western countries operate on a principle of equality before the court regardless of who it is and this principle shall not be broken even if the those in court practice their laws differently. Creating another special system for addressing this would open a way for the abuse of justice thus it better to maintain the system as is.
Murder and copyright violation are different tiers of crime. The path to justice is also much clearer in murders than in the question of "this model trained on mass-pirated data was distilled against the ToS into this other model which was released for free then fine-tuned and resold for dirt cheap."
Most enterprises capable of running that much gear won't take that risk even if it is unenforceable. Oracle has made a great living threatening audits of companies who download its updates.
Yeah, but only Western companies who might have to defend themselves in court. I have no doubt that pirated copies of Oracle DB and the JDK are frequently used in China.
Overusing a piece of software in violation of a TOS/license agreement is radically different than employing human agents to conduct on-the-ground corporate espionage.
After all, _their_ AI is just trained on stolen IP, right? Unlike _ours_?
I still can't get over how irresponsible it is of _them_ to release these weights for free. That means _anyone_ could gain access to their AI model now.
At least _we_ make sure access to _ours_ is fully controlled by _us_.
People laugh, but it is impressive how often this has to be said, and how often it is ignored.
I'm still waiting for the media juggernauts to start beasting on the american hypescalers due to all the theft going around. I know for a fact that both Anthropic and OpenAI had access to some private github repos of mine in the past (due to clear artifacting in their training set). The thing is, they make so much money (allegedly), so they cannot be wrong (unlike the chinese). And this does not take into account all the hacking and system intrusion facilitated (and often perpetrated) by their systems!
Or, you can just not pay and save that money to defend yourself in case they do sue you. Because it isn't established that model weights are copyrightable.
It's a license agreement, not copyright. I give you this thing under certain conditions you agree to. That's not simply about copyright. You agreed in a contract not to do something. Some contracts aren't enforceable but there has to be some major reason. By default people and businesses are free to agree to contract terms they want to.
Pretty sure there are ways to get the model weights without agreeing to a contract. (Indeed, the HuggingFace page appears to link to multiple different ways to acquire the weights without agreeing to any contract)
Wow... I'd like to know more. 40 analysts not spending 100USD/month is 4k/month saved. That means you break even in 27 months. But you probably used Claude much more than that.
Also what harness do you use? Any problems you encountered on the way?
Anthropic is very good at either convincing or forcing (I’m not sure which) companies to pay for tokens rather than for subscriptions. My old job moved from subs to tokens and everyone started running out of budget mid month. It was a bad move for a company that laid off all of QA and 2/3 of Engineering and committed to everyone making up for it with LLMs.
I don't think Anthropic or OpenAI offer subscriptions to enterprises - it's a hobbyist/solo-entrepreneur kind of thing. I seem to recall SpaceXAI had some kind of hack going where they were trying to use subs with Anthropic and then got cut off.
Enterprises also need very stringent IP protections that you don't get through the subscriptions. Even the Anthropic-API per token license isn't great - sometimes you need to go to 3rd parties like AWS Bedrock to get guarantees regarding no multi-tenant GPUs.
Ah - so I was somewhat, but not totally wrong. They do offer subs to SMB (I did not know this!) - but for the Enterprise (typically in thousands to tens of thousands) directed through other channels.
I have only ever paid for tokens and never used subscriptions because I need to know that I can generate sufficient real world value at the true cost of inference.
This seems like setting money on fire. Nothing stops you from tracking token use on your subscriptions and tracking how much you would spend via APIs for if/when you can't use a sub any more, and taking the savings in the meantime.
It's worked out. Definitely not frontier but it's the largest model that works reliably on our hardware. We also try some smaller ones when they come out.
The benefit is that you can use these models on a variety of different providers. You're not stuck with a single proprietary provider. Aggregators like OpenCode Zen and OpenCode Go can shop around providers on these. OpenRouter shows you the costs of a bunch of the providers, and you can route to the cheapest ones.
I regularly use together.ai with models like this, where they're fully self hosting them and I can turn off all data retention, at a fraction of the cost of proprietary models. I recently discovered abliterated models (models edited to not refuse to work on security stuff) and hosts for them like refuseless.com, and that's an option for open source models, not for proprietary ones.
> Aggregators like OpenCode Zen and OpenCode Go can shop around providers on these. OpenRouter shows you the costs of a bunch of the providers, and you can route to the cheapest ones.
Unfortunately due to their licensing requirements for modal-as-a-service providers, the pricing can’t go lower than Moonshot’s from what I can tell.
This is still great if like your example you trust together.ai more, or if different providers have better speeds/cache rates (though usually the original lab has the best cache hit rate).
And obviously if you’re a large enough service to afford to self host or finetune it then this is good.
I didn't dig into Kimi's MaaS legalese, but i dont really understand how that knee-high fence is supposed to work long term.
All the service provider needs to do is provide the option to rent out VM's with access to the inference cards.
Have the customer provide the model.
Maybe provide a bit of a speedup, by letting the customer mount the model by hash.
Sure, that all sounded way too technical 10 years ago, and somebody would have said that's already too much annoyance to deal with for average customers.
Reluctantly i'd agree.
That was 10 years ago. Nowadays, with some proper docs, this entire set up can be up and running in 5 minutes.
In my own experience the vast majority of llm consumers would prefer to pay by token, especially for large (300B+) as either their usage is below the point a dedicated cluster is viable or it’s just a lot more hands on.
With Kimi k3 specifically you need something like 1.6TB of memory which... Well you can’t just use 8x H100’s, and per token pricing becomes extremely attractive at this size for most users.
There's no need to use ablated models right now for cyber work. Several open weight models don't refuse cyber requests (DeepSeek, GLM, Kimi K3, etc). And as you said, OpenRouter has many 3rd party ZDR inference providers and several of those serve these models.
Ablated models are not any more inherently knowledgeable about "forbidden" topics, either. Since such topics are not generally trained for, you just get weird creative hallucinations.
The problem with looking at different prices from different model providers is a lot of the cost is generally driven by cache hit rates.
That’s where you can accumulate expense and the cache hit rate can very wildly from one provider to the next.
You could easily run it with SSD streaming on as little as 192GB RAM (The dense portion is around 115GB or so at native precision, and the remainder would fit KV and a cache for the most commonly referenced experts.) This fits Gorgon Halo, which is a reasonably priced platform. It would be slow, but still usable for high-value tasks.
Excuse me, could you elaborate how you came up with 192GB of RAM? I almost thought they downsized model, but it's still 2.8T tokens. Are we talking about not-yet-out 1-bit quant or assumption that something like ds4 (daskstar, not deepseek) will heavily optimize it?
All these very large models are sparse mixtures of experts (SMoE), where typically only 2-3% of weights are active for a given forward pass.
Usually there are some static experts in each attention layer, plus a whole bunch of dynamically-routed experts. Typically some form of LRU cache is used for the dynamic experts.
So you can keep only the static expert(s) and routing layers in memory and then stream the dynamic experts from SSD, at a modest inference speed cost driven by the latency required to load layers from SSD when they aren’t already in the memory cache.
The LRU cache works reasonably well because a given task will often hit the same experts on each iteration of the forward pass.
For example, on a Mac Studio, Qwen3.8-flash-next (180b total parameters) with 50% of experts in memory and n-gram offloading has about 1/4 to 1/3 the inference speed of everything in memory. That requires 32GB of vRAM for the weights (plus a little extra for context, obviously)
It's not based on loading the full 1.5TB model into VRAM or system RAM, it's just streaming parameters on demand from SSD - which is worthwhile since the MoE portion of the model is quite sparsely accessed. This has been demoed on a 128GB M5 MacBook Pro with lousy performance (0.3 tok/s or so) because it had to load at least ~25GB from SSD per token (comprising the entirety of the sparsely routed active params) and potentially even reload some of the dense weights, depending on what exactly managed to fit on that platform. With 192GB you'd avoid any reloading for the dense weights and even cache a few layer-experts.
Yes, I understand the principle, but will it be usable? E.g. in my personal opinion border of usefulness starts somewhere at 7-10 tok/s. I currently have 192GB and 24GB VRAM, and I can run Qwen-Flash-Next (4 bit) and GLM5.3-Flash (4 bit) at 25 tok/s and 7 tok/s respectfully (with quantized cache on beellama). Parallel queries worsen result significantly but can push aggregate to around 50 tok/s.
I can't, however, run DS4.1-1017, for example, but maybe I didn't find proper options and quantization yet.
So I'm really interested how to be sure what can/cannot fit in specific hardware (for example, I want to buy a Mac Studio with 64 or 128 GB RAM, what MoE model it would run at comfortable speed? Will it be qwen-flash-next or something bigger? And so on).
So far websites with "benchmarks" were really terrible in that regard (and huggingface calculator even worse!)
> in my personal opinion border of usefulness starts somewhere at 7-10 tok/s.
This is expected if you're actively steering the agent, but what about running unattended tasks - perhaps multiple such tasks in parallel, such that you can stay engaged by "attending" to each of them in turn? Then you're not restricted to such a requirement, and can benefit from much slower (individually, not wrt. total throughput) inference streams.
> I currently have 192GB and 24GB VRAM
I'm not familiar with beellama, but this seems like it could be the perfect setup for Strata which people are saying good things about, including for running DS4.1F (at least quantized). The whole point of it is to effectively leverage heterogenous RAM/VRAM setups. Of course llama.cpp itself is doing a good job of trying to catch up, but they've yet to support efficient SSD streaming (mmap doesn't quite cut it) in the main branch.
I run this exact setup with 24GB vram and 288gb ran. I get 25 to 28tps. The catches you have to run the 2.25-bit version currently. Strata is not optimized to run the 3 ir 4-but versions, at least not yet.
With a 256K context window and vision I think it uses around 135 MB— I can’t remember if that’s total or just ram.
Edit: 4.1 flash next
> Strata is not optimized to run the 3 [o]r 4-b[i]t versions, at least not yet.
How so? This is not obvious to me; AIUI Strata does support streaming weights for larger models via SSD. Of course the speed would drop quite a bit once you rely on SSD, but it ought to be workable. Since you're on a HEDT/server platform, you can also benefit by adding more SSD hardware to be accessed in parallel via the abundant PCIe lanes.
> I want to buy a Mac Studio with 64 or 128 GB RAM, what MoE model it would run at comfortable speed? Will it be qwen-flash-next or something bigger?
This depends on how you plan to use the machine... but Qwen-flash-next is right at the top of what a 128GB Apple Silicon can do.
QFN with 4bit quantization will run "comfortably" with all weights in memory and a 128k context on a 128GB machine. On M5 Max w/ 128GB, that will give about 75 tok/sec (using oMLX with MTP enabled). It spits out text much faster than I can read.
However, this truly uses all the RAM. You'll have like 8-12GB left for the OS and apps, which isn't really enough to actually use the machine as a workstation while the model is loaded. But it would work fine as an inference server with Opencode running on your laptop. (Hence, "how you plan to use it?")
Alternatively, you can offload QFN's n-gram table to SSD (this is just a settings option in oMLX), which drops generation speed on the same machine to ~60-65 tok/sec (assuming the model lives on the internal SSD) and reduces memory usage to ~90GB with even a large context. That leaves plenty of RAM to use the Studio as your main workstation. Also, you may run much longer contexts because of the extra memory headroom.
Second alternatively, you can use oMLX's expert offloading feature to keep 50% of experts in memory (with n-gram offloading also). This reduces the RAM footprint to a little over 32GB + context. Offloading the experts drops generation speed to around 20-25 tok/sec on the same machine as above, but it would allow you to comfortably run on a 64GB machine, while still using it as a workstation.
I have tested the 50% expert offload case on an M1 Max / 64GB, and the inference speed is only a little slower (more like 18-20 tok/sec) than M5 Max / 128GB because in both cases the bottleneck is SSD read bandwidth and the GPU is spending a lot of time idle.
One last thing: GLM 5.3-flash doesn't use an n-gram table, so you can't use that trick to reduce memory footprint. You'd need to use expert offloading to make it fit a 128GB machine. Since it has over twice as many active parameters as QFN (which doubles the required SSD read bandwidth), my guess is that a 128GB Studio would not be much faster than your current setup for that model.
Wow, TIL. Is there any calculator or rule of thumb you use to figure out the expected performance of a model on some DRAM/SSD split? For instance I am wondering what the performance could be like on a 512GB M5 Mac Studio.
The rule of thumb is to just test it on your chosen platform, with your workload. SSD performance matters a lot obviously, you'll want to run PCIe 5.0 if at all feasible, or a recent M5 Pro/Max/Ultra or M6 Mac with high storage capacity (lower tiers have much worse performance). You can figure out a worst case easily (assuming that the fraction of sparse params you manage to cache in VRAM or system RAM has a fair chance of being picked among the total) but in practice you do quite a bit better than that - which is why even very small caches can be potentially worthwhile.
I agree, but it would have to be a place with plenty of energy and the only such place outside the US is a war zone.
Perhaps one option is building data centers along with solar power plants in sunny countries that have a lot of space. I'm not sure about the water/cooling situation though.
The rise of China due to hand-over-fist ceeding of power and influence fueled in equal parts by the right and left is so astonishing it'll certainly go down in the history books (written in Mandarin, of course).
Bonus points when the morons are doing so because they get all their political news from a Chinese platform. The [Americans] have gone insane.
The rise of China is due to the tyranny of the markets and the fact that firms laser-focused on quarterly growth buy and liquidate those trying to do responsible, sustained growth. Those that sold their factories were able to swallow domestic competitors who didn’t do so.
The political system just does whatever the captains of industry tell them to do.
A lot of the data center build out has been bipartisan. Unless Dems realize they need to listen to their constituents to avoid a 2028 bloodbath (they probably won't), I wouldn't count on that to save us from RAM/GPU prices.
> a flood of 8x B300 will be homelab ready in a few months
Haha, that's beyond wishful thinking, I'm afraid. We're not even seeing a flood of A100s (6 yo and several generations old). Every GPU is still being squeezed right now, and the trend will probably continue for the foreseeable future.
Last I checked AWS still sells EC2 instances using 2018 era GPUs. Most people underestimate just how much long tail demand there is and how long these cards last, despite the higher power consumption.
Yes, but if data center space gets tight, it creates an incentive to replace old inefficient models with better ones, i.e. sqeueeze more performance out of limited data center space.
IIRC some companies were running campaigns like "we buy your old hardware so you won't sell it in aftermarket and crash the prices". It all ends in landfill.
I wouldn't be surprised if RAM cartel or nvidia will attempt something like that, plus nvidia already talks about license-locking GPUs.
Also, again, IIRC, IBM hardware already works this way. There are almost no aftermarket when I last checked for them because it's all locked up and useless.
Due to costs I am unfortunately years away from ever seeing hardware at this level, but if you just want to study LLM's and AI then there are still plenty of models that work fine on crappy hardware, and you don't have to worry about them becoming sentient on you :-)
A few weeks before Black Friday a couple years ago, I noticed a listing on Amazon for a high-end AI card for around $40,000 (maybe a B100, don't recall). The listing had the automatically inserted Xmas season note "Free returns until Jan 31." I spent a serious minute pondering if I could float the cash for the ~40 days between the credit card bill being due and returning the card. :-)
I assume you mean condition 2 - where they want anyone serving it as a service (to others) and earning more than $20mUSD per year in revenue to contact moonshot and enter a separate commercial agreement.
No, also because that would mean I have to read up on rope scaling or whatever to go beyond the default 256k context, which so far I’ve avoided. But you can allocate it as 220k context for 2 parallel Pi workers, that really speeds things up! Or as ~8x 50k context for dedicated parallel document processing.
Depends on number of users - for a single user with a couple of sessions, you don't quite need that much. Bit of a waste of hardware to only serve this for single users though.
NVIDIA HGX B300 seems to have 8 Blackwells, but I have a hard time getting concrete information from the NVIDIA website on what it actually is, and how you operate it, physically. The picture has this intermediate-product vibe that it should be part of a larger system, like the GB200/300 ones, and I feel like a little boy looking through the window of some shop with incomprehensible things. But I guess the whole point probably is that if you want to have it you put out a tender and someone else will worry about the design and integration and you just pay the bill and point to some empty space in your datacenter cage with large red 5-pole power bundles eagerly waiting.
> Advancing the Frontier of Open Intelligence.
Too bad you probably need something like an 8x B300 node to run this. Black Friday is coming soon though, but I fear this open intelligence will be pacing anyone's liquidity frontier so I'll probably keep trailing the frontier for a while with nvidia/qwen3.8-27B-NVFP4 on dual RTX PRO 4000 in TP=2. I can almost fit 450k context in there.