Xiaomi Open Source Terminal AI Programming Assistant MiMo Code: Exceeding Claude Code in Same Model Score According to Beating monitoring, Xiaomi MiMo has officially released and opened sourced the terminal AI programming assistant MiMo Code V0.1.0, using the loose MIT open source protocol. MiMo Code is based on the open-source project OpenCode for secondary development. It comes with a limited time free MiMo-V2.5 multimodal model and is compatible with mainstream big model APIs such as DeepSeek, Kimi, and GLM, as well as third-party Token Plans. Ordinary programming agents rely heavily on models to autonomously record notes, often forgetting key contexts due to the model not actively triggering them. MiMo Code introduces a persistent memory system that outsources state recording to independent sub agents. When the session window approaches its limit, the sub agent automatically saves the state and reconstructs a clean brief for seamless connection with the main agent. The built-in/stream command runs automatically every 7 days, with an independent agent reading historical sessions and memory files, performing merging, deduplication, and path validity verification, compressing scattered information into the current state to update global memory. To enhance the compatibility between the model and the intelligent agent framework, MiMo Code has designed a dedicated Harness system specifically for the MiMo series models. Users can switch to Compose mode by pressing the Tab key, and after providing basic requirements, the system can independently execute the complete development loop of design, planning, coding, testing, and review. In the authoritative tests of SWE Bench Pro and Terminal Bench 2, MiMo Code using the same base model achieved scores of 62% and 73%, respectively, an improvement of 5 percentage points compared to Claude Code. MiMo Code also has built-in voice control function, which supports users to modify input or execute control actions such as sending through verbal commands. In terms of deployment and usage, macOS and Linux users can install with just one click using the curl command, while Windows users support deployment using npm. After startup, the terminal will present a fully localized TUI interface and a permanent status board on the right side for monitoring work progress at any time. [Original link]