Qwen4-Max Boosts Efficiency with 71B Active Parameters and Cuts API Prices by 25%

·2026.11.17 19:30

Key point

Qwen4-Max improves efficiency with 71B active parameters and reduces API prices by 25%.

1 / 6

Details

  • The Qwen team announced the 2.4T parameter flagship model Qwen4-Max on November 16, 2026
  • Reduced the active parameters of the MoE architecture to 71B, a 25% decrease from the previous version, and applied GDN and QSA to reduce attention costs
  • Reduced QwenCloud API prices by 25% to $1.5 per million input tokens and $4.5 per million output tokens, and expanded the maximum output for thinking mode to 262K
  • Recorded first place in four benchmark items, including 75.7 points on DeepSWE v1.1, but lagged behind competing models in some areas such as ProgramBench
  • Qwen4-Max is provided exclusively via API with weights undisclosed, while only the open-source release of smaller models in the Qwen4 series is planned

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.