Xiaomi MiMo Code tops Claude Code on 200-step tasks
Xiaomi’s open-source MiMo Code claims better results than Claude Code on 200+ step tasks, with free MiMo Auto access to MiMo-V2.5.
Xiaomi's MiMo Code claims stronger results than Claude Code on 200-plus step coding tasks.
Xiaomi has released MiMo Code, an open-source coding assistant tied to its MiMo model family, with claims it can beat Claude Code on tasks that run beyond 200 steps. The launch also bundles MiMo Auto, a zero-configuration access channel that gives developers free, limited-time use of MiMo-V2.5.
| 項目 | 數值 |
|---|---|
| MiMo-V2.5 total parameters | 310B |
| Active parameters per inference | 15B |
| Context window | 1M tokens |
| MiMo-V2.5 price | $0.40 input / $2.00 output per 1M tokens |
| MiMo-V2.5-Pro total parameters | 1.02T |
| MiMo-V2.5-Pro active parameters | 42B |
| MiMo-V2.5-Pro benchmark score | 63.8% on ClawEval |
What changed
Get the latest AI news in your inbox
Weekly picks of model releases, tools, and deep dives — no spam, unsubscribe anytime.
No spam. Unsubscribe at any time.
MiMo Code is built around Xiaomi’s MiMo-V2.5 family, which the company says is optimized for long-horizon agentic coding. The base model uses a sparse mixture-of-experts setup, with 310 billion total parameters but only 15 billion active at inference time.
The bigger 
// Related Articles
- [IND]
A vibe coding workflow keeps AI builds on track
- [IND]
Anthropic’s suspension turns AI release into policy
- [IND]
Helix brings $10B to AI infrastructure buildouts
- [IND]
DOE should turn its land into AI infrastructure fast
- [IND]
OpenAI’s Ona buy adds more reach to Codex
- [IND]
The US should set tokenization rules now, or lose the market