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

· ai llm · Source ↗

Article

TL;DR

Alibaba’s Qwen 3.6 Max Preview rivals Opus on coding benchmarks but benchmarks against Opus 4.5 only.

Key Takeaways

  • Priced $1.3/$7.8 per million tokens vs. Kimi K2.6 at $0.95/$4 — Qwen costs more.
  • Benchmarks cherry-pick Opus 4.5 baseline, ignoring 4.6 which has been out for months.
  • Cloud-only at launch; open-weight Qwen series is the real competitive threat to Western models.

Discussion

Top comments:

  • [seanw265]: Kimi K2.6 scores higher on SWE-Bench and Terminal-Bench at lower price than Qwen
  • [jjice]: Benchmarking against Opus 4.5 when 4.6 exists for months is bad faith comparison
  • [fr3on]: Open-weight Qwen running locally is the real story; cloud Max is for benchmarks

    The interesting question isn’t how Max compares to Opus. It’s how long until the open-weight tier makes the cloud tier irrelevant for most tasks.

  • [wg0]: Chinese providers now closing models and raising prices — open-source era may be ending

Discuss on HN