MiMo-V2.6-Pro: 42B-active open model nears GPT 5.6, runs on 8x RTX PRO 6000

Jasonio · x · 2026-09-22

The newly released MiMo-V2.6-Pro-RL has 1.02T total parameters with only 42B activated, scoring just 1 point below GPT 5.6 Sol Max while fitting on 8x RTX PRO 6000 GPUs. Commenters argue this collapses the cost gap for "frontier intelligence" and explains why Anthropic might want restrictions on open-source AI, making its $2T valuation look shaky. A smaller MiMo-V2.6-Flash-RL (309B/15B) was also released.

Related event: Xiaomi open-sources MiMo-V2.6, tops open-weight rankings(32 posts)→

Original post →

More from Infra

Infra channel →