Xiaomi live-streams MiMo RL training at 2B tokens per step; one user burns 2B tokens in 12 hours
MaziyarPanahi · x · 2026-09-17
Xiaomi turned RL training into a spectator sport: live public metrics for MiMo's reinforcement learning run 2B tokens per RL step, with even VRAM hiccups visible. The author argues more labs should let people watch them build.
After buying the Pro plan, the author's agent logged 2 billion tokens in under 12 hours — already 16% of the monthly allowance — jokingly calling it a stress test of the subscription.
Related event: Xiaomi live-streams MiMo-V2.6 RL training, burning over $1.08M in 36 hours(14 posts)→
More from Models
- Anthropic's server-side rule reignites the case for owning open models over renting AI — alex_verem · 2026-11-02
- Epoch AI: open-weight GLM 5.3 exploited the patch gap to hack 6 Firefox releases at ~$1,000 each — rohanpaul_ai · 2026-11-02
- SemiAnalysis: safety classifiers sandbag scores of Opus 5.5 and GPT-6 Astra — morqon · 2026-11-02
- Liquid AI's 1.7B LFM3 becomes first local model to nail live fixtures lookup on a MacBook — NielsRogge · 2026-11-02
- Claude upgrade-page traffic rebounded to 17.6M in September, near May's 18.6M peak — firstadopter · 2026-11-02
- Open-source GLiNER-PII scrubs 13 identifiers from a discharge note in 550ms on GPU — MaziyarPanahi · 2026-11-02