https://qwen.ai/blog?id=qwen3.8
The model weights will be open-sourced on Hugging Face and ModelScope next week — stay tuned.
I’m hoping they also release a new 35b a3b, for us VRAM poors, a new 9b would also be great!
It’s not clear from the wording whether the 27B model is dropping next week as well, or whether they’re just announcing that there will be a 27B model.
true, it’s not perfectly clear
also I just saw this
I’ve been hoping they would release a refreshed 27b model. 3.6-27b has been my daily workhorse for personal projects.
It has indeed been really good, so much better than similarly sized models a few years ago. Would be interesting to know how far it is from what is possible to fit into a 27b model.
I’m using Gemma4:26b as of right now. Only using it as a conversation agent in home assistant with MCP tool usage.
Has anyone compared gemma and qwen in combination with home assistant and has an opinion?
Gemma is probably good for that, as long as it’s consistently succeeding at the tool calls.
I’d say its above 95% success rate for me personally
Given they have not released 27b or any smaller model with Qwen3.7, I’m not holding my breath for 3.8. They also haven’t said anything public about it.
OP’s image is literally them saying something public about it
No, they said it for 3.7 too, but then never released it, that’s what I meant
Ah that makes sense.
Can’t wait for the new 27b. The old 3.6 is still such a powerful small model. I also agree for the 35b a3b. I thankfully refreshed my PC for LLMs before prices skyrocketed, so I can run comfortably the 27b one Q8, but sometimes having the pure speed of the a3b is very useful.
I hope their decision to opensource Qwen3.8-Max means this version is going to be another open-weights frontier contender 👍
i have a 6700xt with 12gb vram. i need a usable model. 9b is barly usable because of token speed. But non the less i’m exited
How much CPU RAM do you have?
At 12GB you should be using a sparse MoE with offloaded experts. There are a few to choose from, but it just depends on how much spare RAM you have.
The setup described in this Codacus video is what got me started getting over 100k context with decent speeds on my RTX 3070 that I’ve been using daily for the last month: https://m.youtube.com/watch?v=0AqpaFm11oI.
The TheTom fork of llama.cpp adds asymmetric TurboQuant support that allows k at tq4 and v at tq2, which allows squeezing in more KV cache without quality loss. The REAP version of the MoE model also works just fine while further reducing the model size.
The TheTom fork doesn’t have pre-built container images, so it’s necessary to build your own (the ROCm Dockerfile in the devops directory).
If anyone does want to use the fork, I just recently built the ROCm image myself and ran into an issue where the fork’s Dockerfile was pulling a UI build package from Huggingface that doesn’t exist anymore, so I swapped in the latest ROCm Dockerfile from upstream that now builds the UI from source, and that worked fine.
Yeah do what these guys are suggesting and do the MOE with the experts offset. It really works well.
have you tried Qwen 3.6 35b a3b? check my guide, it’s still relevant to you just with different numbers because you have 12GB
I only have 8gb vram and can run 35b-a3b perfectly (30+ tps), have you tried 35b?
With which quantization ? I have a 16GB GPU, running Qwen3.5 9b UD_Q8_K_XL basically max out the VRAM usage. Maybe a MOE model fits you. Gemma 4 e4b is one with a total 8b.
And what is the token speed you are getting ?
