QwenAnalysisAI ModelsAugust 26, 2026

Qwen 3.8 27B is excellent, but it defaults to wildly overthinking things

Alibaba's Qwen 3.8 27B, an Apache 2 licensed 27B parameter vision-capable LLM, ships with a default reasoning effort of 'xhigh' that causes spectacular over-thinking. Simon Willison found it excellent but recommends raising the context limit to 262,144 tokens to avoid running out.

How this story unfolded

4 weeks · 17 reports · 173 community posts · 190 of 204 shown

  1. Aug 3
  2. Aug 11
  3. Aug 12
  4. Aug 13
  5. Aug 14
  6. Aug 15
  7. Aug 16
  8. Aug 17
  9. Aug 18
  10. Aug 19
  11. Aug 20
  12. Aug 21
  13. Aug 22
  14. Aug 23
  15. Aug 24
  16. Aug 25
  17. Aug 26
  18. Aug 27
  19. Aug 28
  20. Aug 29
  21. Aug 30

Qwen by email

Get an email when Qwen has news

No news that day, no email.

More stories today

Open the live feed