1. Deepseek V4 is still in preview (training is not finished)
2. Qwen is much more demanding and borderline unusable on consumer hardware because it's a dense model. The 27B parameters are active all time for each token. It's not a MoE architecture where a router activates only some of them.
I have to disagree with most claims. I run Qwen3.6-27b at 260k context and 40-60 tok/sec. It handles most coding problems as well as Sonnet 4.6 under OpenCode on our production tasks. (As an experiment, I run the same prompts for the same issues in parallel for Qwen 3.6 and Sonnet 4.6 and usually see little difference in performance). I see zero degradation from quantization in practice.
Last time I tried running large MoEs on this PC, they had inferior quality at 2-3 bits compared to much smaller dense models at 5-6 bits, and were slower anyway.
A 260k context (close to the stock maximum for Qwen, though it's possible to extend it) will take ~16GB RAM for storing the KV cache, barring quantization tricks which severely degrade quality. That's a whole lot more than what DeepSeek requires for a similar context length, and makes it infeasible to batch multiple inferences together. This used to be the status quo for consumer inference, in fact it still is for models like Kimi and GLM (which can sometimes be smarter than even DeepSeek V4 Pro!) but we can also do better nowadays.
Deepseek V4 Flash still has 13B active params though? That is about half as many as Qwen3.6-27B (and much more than Qwen3.6-35B-A3B). Given that RAM (even on a base M4 or 'regular' Intel/AMD system) is like an order of magnitude faster than an SSD, even Qwen 27B running from RAM will be much faster than any Deepseek V4 model with SSD offloading. And the MoE will be much faster still.
Qwen 27B is also small enough to completely fit in a high-end consumer or mid-end pro GPU, like an RTX 5090 or Radeon PRO R9700. I found results claiming 30 tokens per second generation for 27B(-Q4_K_XL) on an R9700. I doubt you get more than 5 tokens per second doing SSD MoE streaming.
Even for relatively short contexts, I honestly already find the ~30B class MoE models to be only borderline acceptable in terms of speed on my laptop (Ryzen 7 7840U, 64 GB LPDDR5-6400), though I use Gemma 4 26B-A4B more than Qwen3.6 35B-A3B.
> even Qwen 27B running from RAM will be much faster than any Deepseek V4 model with SSD offloading.
If you have reasonable amounts of RAM to cache the most likely experts, that's not true at all. Qwen 27B is marginally faster on a nearly empty context, then falls behind as context length increases due to the different attention mechanisms. Prefill for Qwen is much faster, but you're still comparing vastly different model sizes and capabilities. DeepSeek Flash is the best deal overall.
> completely fit in a high-end consumer or mid-end pro GPU
Or you could fit the dense portion of a much more capable model and still take advantage of that hardware.
Is that how MoEs work? I though that an important constraint for MoEs is that experts need to be uniformly used to make sure they can be used effectively. If there is a 'common subset' that, if anything, sounds like a symptom of undertraining (i.e. the same trick will not work as well for Deepseek V4.1).
Also, even if your MoE hitrate is 90%, you still spend half your time waiting for the SSD, giving similar total speed to a 27B model!
Finally, it looks like Deepseek V4 is pretty much only runnable with antirez's ds4, and SSD streaming only works with Metal; but I would like to try what you say with llama.cpp which uses mmap to also potentially do SSD streaming. (I can maybe try the large Qwen3.5 MoEs?)
> as context length increases
What kind of context length do you consider reasonable, though? From what I know, all models (even frontier ones) start degrading once you pass a few hundred thousand tokens. So realistically, limiting context size might even improve quality, especially if you use token-efficient harnesses.
> Or you could fit the dense portion of a much more capable model and still take advantage of that hardware.
Your point about consumer hardware was that it would be "borderline unusable" when running Qwen 3.6 27B. However, you need much less hardware to run a 27B than DSv4 Flash. In addition, you can do the same 'trick' with low-end GPUs and small MoEs: my desktop with 32 GB DDR4-3200 and an RTX 2070 8GB can run the ~30B class MoEs at 20-30 tokens per second and similar speeds to my laptop.
For any given workload/session? Empirically, yes, that's what has been found across different models. There's quite a bit of predictability that makes caching helpful.
> Also, even if your MoE hitrate is 90%, you still spend half your time waiting for the SSD, giving similar total speed to a 27B model!
There are ways of masking some of that latency, though it requires some architecture-specific cleverness which is less directly applicable to a generic engine like llama.cpp.
> Finally, it looks like Deepseek V4 is pretty much only runnable with antirez's ds4, and SSD streaming only works with Metal
The llama.cpp folks are working on adding support, and the ds4 project is working on CUDA support for streaming inference, targeting the DGX Spark.
> From what I know, all models (even frontier ones) start degrading once you pass a few hundred thousand tokens.
DeepSeek V4 seems to do quite well on recall tasks even with large context. That's one plausible benefit of its compressed attention mechanism, compared to earlier models. Some degradation will likely still be there, but it's not necessarily obvious.
As for why people are calling Qwen 27B "borderline unusable" that may have to do with it being a dense model which makes for an increased compute intensity and pushes users towards discrete GPU platforms, since those tend to have the most compute overall as far as consumer hardware is concerned. I might agree that Qwen 27B is quite ideally tailored towards these platforms, but that does come with some limitations.
2. Qwen is much more demanding and borderline unusable on consumer hardware because it's a dense model. The 27B parameters are active all time for each token. It's not a MoE architecture where a router activates only some of them.
3. Qwen doesn't like quantization at all.