Qwen 3.8 Flash Next
Vision enabled, still going strong :) It's FAST. 9s - Image description 3m09s - 3d Visualisation Take a tour :) 4x3090, VLLM, 256k ctx Average: ~100t/s albucino/Qwen3.8-Flash-Next-W4A16-FP8PLE
Your security co-pilot: recon ,reverse engineering and threat intel in one agent , on your machine not someone’s cloud
See it , break it . Report it.🧠 We support Qwen 3.8 flash Qwen 3.8 max GLM 5.2 And GLM 5.3 all of our models are uncensored
Alibaba's Qwen 3.8 Max isn't just reproducing research papers—it's self-evolving.
Starting from scratch, it reproduced results and even tested 18 of its own improvement ideas. This recursive self-improvement is a step towards AI making its own discoveries.
The surprising part: Qwen 3.8 Flash Next is being tried locally in DwarfStar on Apple Silicon, with a claimed 65 t/s on an M3 Ultra.
The test ask is specific: 64GB Mac, Q2 weights on HF, PR branch on GitHub.
You have reached the end of the archive
All of qwen38