Xiaomi Open-Sources MiMo Code, Claims It Outperforms Claude on Long-Horizon Coding Tasks
Summary
Xiaomi open-sources MiMo Code V0.1.0, a terminal-native AI coding assistant claiming to outperform Claude Code on long-horizon tasks, featuring a cross-session memory system, one-million-token context window, and aggressively low pricing at $0.40 per million input tokens — though self-reported benchmarks and early-stage status warrant cautious optimism.
Key Points
- Xiaomi has open-sourced MiMo Code V0.1.0, a terminal-native AI coding assistant that outperforms Anthropic's Claude Code on key agentic benchmarks, particularly on long-horizon tasks exceeding 200 steps, based on an internal beta involving 576 developers.
- MiMo Code tackles AI coding agent memory loss through a cross-session SQLite-powered memory system with four layers and an independent checkpoint-writer subagent, enabling the tool to rebuild context from structured checkpoints without losing operational momentum.
- Available under an MIT license with free limited-time access to Xiaomi's MiMo-V2.5 model featuring a one-million-token context window, MiMo Code enters a crowded field of terminal coding agents with aggressively low pricing starting at $0.40 per million input tokens, though its V0.1.0 status and self-reported benchmarks signal early-stage maturity.