Hands-On with Qwen3.8 Preview
葬AI · wechat · 2026-07-19
Based on the ZangAI benchmark, this quick review compares several models: Fable5 leads the pack, while qwen3.8-max-preview and KimiK3 are roughly in the same tier. However, KimiK3 shows more volatility with more low-score rounds, which the author speculates might be due to insufficient inference resources.
The author also points out that while qwen3.8-max-preview's coding capabilities are on par with KimiK3, its non-coding abilities are noticeably weaker. In tasks like image/3D generation, it falls behind GPT-5.6-Sol, KimiK3, and Fable-5. As this is just a preview version, the author expects further updates and tweaks before the official release.
Related event: Kimi K3 vs Qwen 3.8 Max: Fierce Competition in Chinese LLMs(4 posts)→
More from Models
- MissionReport benchmark: only 3 of 540 AI-generated government reports pass all criteria — Exp_Mark · 2026-10-27
- Local beats cloud: Qwen 3.8 FN handles private student data as well as frontier models — dr_lm · 2026-10-27
- Aidan Gomez: model's CoT said no tool calls right before making them — aidangomez · 2026-10-27
- Nat Lambert: Cohere's model is narrow RL training on internal infra ill-suited for open weights — natolambert · 2026-10-27
- OpenAI teases DOTS is coming to Europe 'soon' — mark_k · 2026-10-27
- Clifford Pickover challenges AI to decipher a message he encoded in Norse runes — pickover · 2026-10-27