Alibaba releases Qwen3.8-27B, a local multimodal model

How this story unfolded

4 weeks · 6 reports · 43 community posts · 49 of 52 shown

1. Aug 3
   • 2 community posts
2. Aug 14
   • unsloth/Qwen3.8-27B-NVFP4huggingface.co
   • 8 community posts
3. Aug 15
   • 3 community posts
4. Aug 16
   • Qwen 3.8 27B is excellent, but it defaults to wildly overthinking thingssimonwillison.net
   • 2 community posts
5. Aug 17
   • 3 community posts
6. Aug 18
   • Qwen3.8-27B runs frontier-class coding agents and reasoning locally, no cloud API requiredventurebeat.com
   • orcarouter/Qwen3.8-27B-Uncensored-MLXhuggingface.co
   • 5 community posts
7. Aug 19
   • 2 community posts
8. Aug 20
   • 2 community posts
9. Aug 21
   • Qwen 3.8 Low and Medium are goated
10. Aug 22
    • I benchmark DFlash 2 (PR build) in llama.cpp on Qwen 3.8 27B against all speculative methods for 3 days. 2.26x on 100 real coding prompts, 4.68x with one n-gram drafter on top. Up to 8x on specific cases.
11. Aug 23
    • outsourc-e/Qwen3.8-27B-Unleashed-GGUFhuggingface.co
    • 3 community posts
12. Aug 24
    • 4 community posts
13. Aug 25
    • Qwen3.8 27B is now available for both fine-tuning and Dedicated Model Inference Fine-tune it on...
14. Aug 26
    • Alibaba Releases Smaller, Cost-Effective Qwen AI Modelbloomberg.com
    • 2 community posts
15. Aug 28
    • 3 community posts
16. Aug 30
    • Here my pretty good qwen3.8 27B setup, hope it helps