Xiaomi's MiMo-V2.6 lands: 1.02T-param MoE takes top open model spot on Artificial Analysis
_AndrewZhao · x · 2026-09-22
Nathan Lambert highlighted Xiaomi's new open MiMo-V2.6 models: MiMo-V2.6-Pro (1.02T total / 42B active) and MiMo-V2.6-Flash (310B / 15B active), with Pro topping Artificial Analysis' open model leaderboard. The release includes an RL dashboard and tech report — the second paper on MiMo's RL run, which observers call evidence that self-improvement has gone mainstream.
More from Models
- Grok 4.7 Slammed in Benchmarks: 2.5x Tokens for Just 2 Points More — ivan_bezdomny · 2026-09-22
- Claude's 20x usage add-on appears purchasable again after lockout — CynicalCandyCanes · 2026-09-22
- JevBench v1.2.16 update: Winnow-12B debuts at #5 — airesearch12 · 2026-09-22
- Grok 4.7 tested: HTML water glass and procedural watch exploded views impress — Daniel_Farinax · 2026-09-22
- ChatGPT skill sync between Chat and Work broken, users suspect dark UX to burn usage — jdjohnson · 2026-09-22
- Laya runs at 84ms but loses every game: fast local models aren't smarter — eter_inquirer · 2026-09-22