Qwen3.6-Max-Preview: Smarter, Sharper, Still Evolving

· top-stories ai llm · Source ↗

Article

TL;DR: Alibaba’s Qwen3.6-Max-Preview is a strong cloud-only SOTA competitor, pricier than Kimi K2.6.

Key Takeaways

  • $1.3 in/$7.8 out vs Kimi K2.6 at $0.95 in/$4 out; Kimi slightly better on SWE-Bench Pro
  • Benchmarks compare against Opus 4.5, not 4.6 — credibility questioned across thread
  • Chinese providers shifting from open weights to closed cloud-only models, raising prices

Discussion

  • Chinese models increasingly competitive for coding vs Anthropic at a fraction of the price
  • Concern: China following Western labs in closing weights and jacking up pricing simultaneously
  • Local open-weight Qwen 32B/72B still strong; gap to hosted Max narrowing each release

Top comments:

  • [seanw265]: Kimi K2.6 released same day — cheaper and scores slightly higher on SWE-Bench Pro
  • [fr3on]: Open-weight Qwen matters more than the hosted Max; gap is shrinking every release
  • [wg0]: Chinese providers now closing source and raising prices 100% — pattern shift
  • [atilimcetin]: Qwen outperformed Claude and Gemini on specific technical math and graphics tasks

Discuss on HN