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 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).
MoE is a concept proposed in 1991, before the deep learning era (which is before what I call the transformers era). You can think of it like sharing.
Contrary to popular belief; 'experts' in MoE LLMs do not specialize. There's no expert trained to be good at maths, or python, or writing, or whatever. It's an inference optimization.
Wait I thought the router ends up specializing the experts?
Like there is no explicit goal aside from each 'expert' getting roughly equal weight?
And it happens that when you train the router you do end up passing certain classes of problem to each expert - just as a training result nothing as clean as a python expert. But math vs creative writing will tend to rely on different experts over the majority of the inference?
I do not know what I am talking about, this is my limited understanding...
My understanding of it is also pretty surface level, but I was not under the impression that it develops "expertise" in a particular subject matter, at least not in a way that's easy to harness. From what I've read, it develops expertise at the token level.
Because the natural continuation of this is to say like, "Ok I want to load the bird detection expert and the navigation expert but leave the medieval European history expert behind", and my understanding is that this is not really how it works. At least at the moment.
You are right that at training the main goal is balancing between the sections to avoid certain paths becoming the only path. In the end the inference will be routed token by token to a mixture of say 3 or 4 sections of the model. The combination can change at each turn. It’s really a statistical optimization.
As for many things in neural networks, the original intuition coming from anthropomorphism once implemented becomes something very non human!
yes, I average 80-120 tok/s on my RTX 3080 with gemma 4 and faster with Qwen 3.5. The main use-case here is just code-monkey agents. I'm not looking for architectural guidance, but an agent to take a spec and complete it.
> 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.
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.
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
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 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.
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.
So, instead of having one or two maintained projects that work super well, now we have hundreds or thousands of people half-assing it locally each time they need it? Doesn't sound efficient.
you're right, but flipped on its' head it also means that a well-meaning FOSS project may not ship out a zero-day in the future that affects half the planet at once.
now we'll just always be vulnerable from different angles in different and unique ways across the globe.
better? well, no.. but it is different, and biology has offered tons of wisdom about why it's a good thing to achieve things different ways across the world, maybe some far-future version of this weird fragmentation of work that is happening now will contribute to some kind of herd effect that reduces the severity or magnitude of some bug or work of malice.
who knows though, i'm probably just considering it too positively. The disintegration of consolidated effort is a pita, I agree.
There has to be a middle ground between a monoculture and not sharing a single knowledge bit between every human. Because if everyone is making their half assed thing just for themselves using LLM without sharing anything, that also means FOSS will stop existing, and knowledge will only evolve inside the corporations that have money for creating their provate knowledge bases.
It might make sense to do this if there literally isn't an existing solution out there, but I think most would find that it's better to use an existing solution than to recreate the wheel every time there's a need. It's just less effort/time/money.
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.
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.
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...
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)?
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?
IIUC, Kimi K3 on RTX 6000 Ada (48GB) takes 292 s/token
https://github.com/lyogavin/airllm/releases/tag/v3.1.0
Weeps...
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).
MoE is a concept proposed in 1991, before the deep learning era (which is before what I call the transformers era). You can think of it like sharing.
Contrary to popular belief; 'experts' in MoE LLMs do not specialize. There's no expert trained to be good at maths, or python, or writing, or whatever. It's an inference optimization.
As for reasoning in non-text modalities, you might find this paper interesting :) https://huggingface.co/papers/2502.05171
Like there is no explicit goal aside from each 'expert' getting roughly equal weight?
And it happens that when you train the router you do end up passing certain classes of problem to each expert - just as a training result nothing as clean as a python expert. But math vs creative writing will tend to rely on different experts over the majority of the inference?
I do not know what I am talking about, this is my limited understanding...
Because the natural continuation of this is to say like, "Ok I want to load the bird detection expert and the navigation expert but leave the medieval European history expert behind", and my understanding is that this is not really how it works. At least at the moment.
was this meant to read "sharding"?
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.
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.
If you had a deadly condition, and no diagnosis worked, and a specific model had the answer... past that I wouldn't use it.
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 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.
Hoping a winner emerges with some real momentum behind it.
now we'll just always be vulnerable from different angles in different and unique ways across the globe.
better? well, no.. but it is different, and biology has offered tons of wisdom about why it's a good thing to achieve things different ways across the world, maybe some far-future version of this weird fragmentation of work that is happening now will contribute to some kind of herd effect that reduces the severity or magnitude of some bug or work of malice.
who knows though, i'm probably just considering it too positively. The disintegration of consolidated effort is a pita, I agree.
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?
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.
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...
It seems like this tool saves on both disk space and RAM, then. Classic trade off: speed vs space.