Hacker News
Qwen3.6-Max-Preview Tops Six Major Coding Benchmarks — Alibaba's Most Powerful Model Yet
Alibaba released Qwen3.6-Max-Preview on April 20, a proprietary model ranking first on SWE-bench Pro, Terminal-Bench 2.0, SkillsBench, QwenClawBench, QwenWebBench, and SciCode. It shows gains of +9.9 on SkillsBench and +10.8 on SciCode vs predecessor Qwen 3.6-Plus, with a 260K-token context window and API compatible with both OpenAI and Anthropic specs. The 630-point HN story (335 comments) arrives back-to-back with Kimi K2.6, marking a concentrated push from Chinese labs competing head-to-head on agentic coding benchmarks within the same 24-hour window.
↳ Follow the thread