▲ 232 points
back
84 comments
For anyone wondering “how slow is this?”
IIUC, Kimi K3 on RTX 6000 Ada (48GB) takes 292 s/token
I hope I'm not the only one who misread it as 292 tokens/s and got excited momentarily
I thought I met some huge breakthrough on computer science until I saw your comment
haha its something i look out for now. having seen a few recent stories like this.
Most definitely not.
Weeps...
that's 0.003 tokens/second. To get an hour's work done that's normally 30 tokens/second (108k output tokens in an hour) will take 416 days at this rate. And if you're using 100 watts, during that time you will spend $124.61 in electricity, as well as not being able to use your device for something else, plus the noise and heat from your device.
For $124, on Moonshot's official Kimi K3 API rates ($0.30 per 1M cached input, $3 per 1M fresh input, $15 per 1M fresh output), you can purchase 42 million fresh-input tokens, or 8.3 million generated output tokens, in whatever mix you want.
So what you get is 80x more expensive and you wait 416 days to get it.
I do think these “run a bigger model than will fit in VRAM” projects are necessary steps, but are they functionally useful or helpful to anyone currently? For example, is anyone out there running a big Qwen for coding on a 16-32GB machine with these techniques?
> but are they functionally useful or helpful to anyone currently?
Yes and no, depend on your expectations. Some/many like to run local LLMs just for the sake of it, so anything will do.
MoE are useful on PC systems, at the condition of having high enough memory bandwidth (and large amounts of RAM) - that is, Threadripper/Pro.
The advantage of MoE is that only a subset of the model's experts is used for each token, so not all weights need to be present in VRAM at once. The remaining weights can reside in system RAM, although moving and accessing them still carries a substantial performance cost (and that's why high memory bandwidth is needed).
If I could justify wear and tear and electricity, I was willing to do something like this for batch processing. The batches would be a bunch of prompts whose outputs I'd look at the next day. Maybe common operations, like QA or refactoring, on whatever software I wrote.
If so, I could use a larger model than I have real-time hardware for. The largest, well-trained models can often get the output mostly right in one try. I also would be using AI's as a supplement to, not replacement for, my own brain. So, issues with the outputs wouldn't be a problem because I'm just keeping what's helpful.
If I still need to re-generate it all, it might still save money over time by avoiding cloud costs. Also, hardware that's already paid for is a sunk cost that doesn't inflate over time. Glitches in loading or destroying VM's might blow up into a big bill.
> but are they functionally useful or helpful to anyone currently?
I've been building a SaaS that deals with data that can't be distributed to third parties. Some of the useful AI stuff I can add is not time sensitive and can run overnight. Things like this allow me to use higher quality models without selling my house for GPUs.
Wow, you could do a lot at 292 tokens a sec—oh.
I have all praise for those taking this on and in my idiom would call it *the lord's work."
The image I reliably summon to mind is that compilation video showing the progress of Boston Dynamics bots. The curve between technically functional, to comically slow, to too slow for "real" work, on to, OMFG, may prove a (rough) curve.
It's work like this that moves things forward.
Ahaha thank you, I naively assumed the unlabeled graph in the readme was tps, not spt!
Running 292 tps on K3 would make your gpu a money printer.
At that point, how does this compare with simply running the model on the CPU?
Not an answer to your question, but maybe this has more info? I think with these optimizations and quants, it compares negatively. But can these optimizations be applied to models you want? Another question.
https://news.ycombinator.com/item?id=48353348
if you take a model that requires 200GB of VRAM and you run it on the CPU, it requires 200GB of RAM instead. Still unfeasible on consumer hardware. With this approach you can easily do it on 12GB or less of either RAM of VRAM, at several seconds per token instead of tokens per seconds. Very unusable, but certainly interesting!
I have see a lot of "run <gigantic llm model name here> using cpu" or using 1gb of ram... and the short answer is sure, you can run it in 1mb of ram, or in a 286, it will just take a couple of years to produce the first token.
the canary in the coal that this has not change is nvidia share price.
I wonder what this measures in J/token.
Assuming 30% gpu power utilization because of all the loading and unloading 29.2 kJ per token
How many is that in tokens per Scaramucci?
It matches my coding speed...its ok.
Hah I was looking for it and couldn't work out how many years/token. 292s is pretty good.
Are there benchmarks for smaller models?
Seeing a lot of these “run 1TB models with 1GB RAM” projects recently. Most seem vibe coded and probably won’t be maintained.
Hoping a winner emerges with some real momentum behind it.
No no, don't just say "vibe coded", say "Fable and $500 of credits"
How a patchwork of Python modules constitutes an API(?) or whatever(?) and how to use it (does anyone?), is beyond me. There is nothing that I would call documentation, let alone concrete usage examples. I know more amusing ways to waste my time. When I want to play with models, I use llama.cpp.
You don't really need a maintainer when codex or claude code can set it up for you; thats how I got Trellis2 working on windows and tiny VRAM despite Microsoft recommending you have 24GB VRAM and Linux. Models are pretty disposable now.
I love how goal posts are shifting from "vibe coded apps dont really work" to "vide coded projects wont be maintained"
I've been seeing a lot of projects that let one use large models on machines with small amounts of memory. They seem to all be doing significant quantization and/or expert streaming. What's the benefit of these projects over something like taking an unsloth quant and running llama.cpp with appropriate flags (-cmoe/-mmap) to manage VRAM vs RAM vs SSD?
I love it how the rampocalypse is pushing people to squash all the performance they can. I hope that this also leads to rethinking model architecture so that they can do the same with less.
I’m still slightly confused on what this adds.
Let’s say I wanted to run a full size open weight model. I have a 128GB m3 max laptop.
Does this basically load layers in and out on demand? So I still have to download the full model to disk, but the RAM requirements go way down? The readme calls out that one still needs to connect HuggingFace, which leads me to believe that maybe you don’t even need to download the full model?
If you point it at a huggingface model identifier, it will download it, I assume. No way around that.
It reads like it is keeping only the core and the active layer loaded at any one point, and streams layers from disk; there are several other solutions like this and if my understanding is right, this is probably better than an mmap implementation or just streaming experts in.
Not an expert in this field, but the "expert" is consisting of multiple layers. To keep it small in terms of memory print, this project streams each layer (dividing even further).
It also requires extra space because of decomposition of the layers. Normally the file format optimized for compute intense workloads. But here the bottleneck is the memory capacity.
Also guessing that you need to be able to hold at least 3-layers at once in the memory, given M x N = R operation, M is the previous layer, N is next, and R is the result. on the next "layer", the R (result) becomes M, gets computed against the next layer, N, yielding the further result R'. And so on, until all layers are processed.
I assume it's horribly slow, but can be put in a non-intrusive background task...
Obviously needs downloading eventually :).
It seems like this tool saves on both disk space and RAM, then. Classic trade off: speed vs space.
I guess the use case is something like: you have a slightly obsolete Mac or PC or a whole bunch of them, and just need to compose one or more convincing spam emails, but it's fine if it takes a full week to do it?
And you don’t pay for the electricity.
You can run any frontier model on your PC if you just wait long enough...
What device do I need and how much will it cost to install one at home so that it works as quickly as the Claude Code answer (and it answers quite slowly)?
If you want it to respond in under a minute, you need more hardware than this application is intended for. This thing's response is measured in seconds per token, not tokens per second.
To get Claude Code responsiveness from even a pretty small (but still usable) model, you need, maybe two 32GB GPUs? I run Gemma 4 31B and Qwen 3.6 27B on my dual 32GB GPU setup (cheap old Radeon Pro V620 GPUs) at about 20 t/s, which is not fast enough for comfortable interactive agentic use. A couple of new GPUs, like Radeon AI Pro 9700 at $1400 each, probably gets you fast enough for comfortable interactive use with small models like those. Those small models are not competitive with Claude models, however (maybe they beat Haiku sometimes). They can write a little Python or make a web page, they can't architect a real application.
To run a near-frontier model like Kimi K3 or GLM 5.2 at comfortable speeds, you need serious hardware with 768GB VRAM, minimum. I think Asus is releasing something like that for about $150k soon. You can run DeepSeek V4 Flash at almost comfortable speeds and in a decently capable quantization on two DGX Sparks or Asus GX10s (about $10,000).
Or, you could use DeepSeek V4 Flash from DeepSeek.com, at blistering speeds and with huge contexts, for something like a decade or two for that same $10,000.
We will see if this project has legs. This is the kind of efficiency we desperately need. Now if we can address efficiency with llm training.
“Can run” and “is useful interactively” are different benchmarks. At this latency, I can still imagine batch or overnight jobs being interesting; for chat, time to first useful answer matters much more than whether the weights technically fit. A workload/latency/energy table would make projects like this easier to evaluate.
"The layer-by-layer streaming is clever.
Curious how the throughput compares to running a quantized model on the same
GPU — seems like quantization might still win on speed?"