FULL STORY
Xiaomi's MiMo-V2.6: Open-Source RL Stack to Top of the Leaderboard
Xiaomi released MiMo-V2.6 with a fully open RL training stack, then its Pro-RL model topped the Artificial Analysis open-model leaderboard at a training cost of just $2.6 million, all under the MIT license.
2026-09-30 ~ 2026-10-03 · 2 episodes · 4 posts
Episode 1 · Xiaomi Open-Sources Full RL Stack for MiMo-V2.6 (2026-09-30, 2 posts)
Xiaomi released MiMo-V2.6 on Hugging Face along with its complete open-source RL training stack, including 7.8K training environments, a public training dashboard, a detailed post-training report, and browser-based rollout tools.
- Xiaomi open-sources MiMo-V2.6 RL stack with 7.8K environments and live training dashboard — SergioPaniego · 2026-09-30
- All Xiaomi MiMo-V2.6 RL resources: Explorer, Harbor, live dashboard, tech report — SergioPaniego · 2026-09-30
Episode 2 · Xiaomi's MiMo-V2.6 Tops Open-Source Model Rankings with $2.6M Training Cost (2026-10-01, 2 posts)
Xiaomi released its fully MIT-licensed MiMo-V2.6 models, with MiMo-V2.6-Pro-RL topping Artificial Analysis' open-source intelligence leaderboard. The technical report reveals a 'You Only RL Once' approach using mixed GRPO training across domains, with training costs of about $2.6 million.
- Xiaomi MiMo-V2.6 report: one mixed GRPO run across all domains, $2.6M for Pro — SergioPaniego · 2026-10-01
- Xiaomi's MIT-licensed MiMo-V2.6-Pro-RL tops open-weights intelligence index, discloses ~$2.6M RL training cost — lmoroney · 2026-10-03