Alibaba releases Qwen3.8-Max to compete with western AI

Short summary

Alibaba launched Qwen3.8-Max, a 2.4T-parameter mixture-of-experts model activating 95B parameters per inference, targeting enterprise software engineering and complex reasoning. Open-weight versions will be available via Alibaba's cloud studio platform. The company claims competitive performance against Claude Opus 4.8 and GPT-5.6 Sol on coding benchmarks, and reports unsupervised coding runs lasting up to 16 days. Critics question the details of these autonomous claims.

• •Qwen3.8-Max: 2.4T MoE model, 95B active params, open-weight release targeting enterprise coding and reasoning
• •Alibaba claims parity with Claude Opus 4.8 and GPT-5.6 Sol on SWE-bench Pro coding benchmarks
• •Company reports 16-day unsupervised coding project completion; critics question autonomous run details

Generated with AI, which can make mistakes.