PyyanThe AI Index

Pyyan / Compare / Qwen3-235B-A22B vs MiMo-V2.6-Pro

Qwen3-235B-A22B vs MiMo-V2.6-Pro

2 of 5

Open-Weight Models · verified 29 Sept 2026

Qwen3-235B-A22BAlibabacurrent
XMiMo-V2.6-ProXiaomicurrent
3 slots left
SpecificationQwen3-235B-A22BMiMo-V2.6-Pro
SummaryCheapest capable model, most permissive licence.A trillion parameters under the MIT licence.
Context—1M
Input ($/Mtok)—Weights only
Output ($/Mtok)—Weights only
LicenceApache-2.0MIT
Max output—Not published
Parameters235B MoE1.02T total, 42B active
Released—21 Sep 2026
Model ID—XiaomiMiMo/MiMo-V2.6-Pro
CategoryOpen-Weight ModelsOpen-Weight Models
OfficialAlibaba ↗Xiaomi ↗

Highlighted rows are where these differ.

Qwen3-235B-A22B

  • No licence conditions whatsoever
  • 22B active of 235B total

Best for commercial self-hosting.

Full spec sheet →

MiMo-V2.6-Pro

  • 1.02T total parameters, 42B active per token, a 4.1% activation ratio
  • Text, image, video and audio in, with a 1M token context
  • Shipped with 7,000+ reinforcement learning environments and the framework that trained it

Best for multimodal work on your own cluster.

Full spec sheet →