MiMo Code 0.1.0 Released: Model‑Agent Collaboration Drives Self‑Evolving AI Coding

MiMo Code 0.1.0, an open‑source terminal AI coding assistant built on OpenCode, introduces a persistent memory system, a Compose mode that orchestrates design‑to‑test workflows, voice input, and benchmark‑proven superiority over Claude Code on SWE‑Bench and Terminal Bench, while supporting multiple large‑model APIs.

Xiaomi Tech
Xiaomi Tech
Xiaomi Tech
MiMo Code 0.1.0 Released: Model‑Agent Collaboration Drives Self‑Evolving AI Coding

Release Overview

MiMo Code V0.1.0 is officially released and open‑sourced under the MIT license. It runs in the terminal as an exploratory AI programming assistant and is built on the OpenCode project.

Core Capabilities

Persistent memory + unlimited context : The assistant uses a three‑layer memory mechanism—project memory, session checkpoints, and task‑progress tracking—to avoid the “AI forgetfulness” problem even after hundreds of dialogue rounds.

Engineered sub‑agents : Instead of relying on the main model to take notes, a dedicated sub‑agent automatically saves state and rebuilds a clean brief when the window is near capacity, allowing the main agent to continue work without restarting.

Compose mode : Press Tab to switch to Compose mode, provide a brief idea, and the system automatically performs design, planning, coding, testing, and review, delivering an industrial‑grade code package.

Benchmark Comparison

Two real‑world coding test suites were used: SWE‑Bench and Terminal Bench . With the same MiMo‑V2.5 model, MiMo Code achieved 62% on SWE‑Bench Pro (Claude Code 57%) and 73% on Terminal Bench 2 (Claude Code 68%), demonstrating the advantage of the dedicated Harness and Compose mechanisms.

Voice Input and Memory Consolidation

The tool includes voice input powered by MiMo‑V2.5‑ASR. Users can speak commands, correct prompts, or issue actions like “send” or “execute” without touching the keyboard.

A special /dream command runs every seven days, invoking an independent agent that merges, deduplicates, validates, and compresses past sessions into a compact global memory snapshot, enabling the assistant to retain knowledge across uses.

Installation & Usage

Mac/Linux : curl -fsSL https://mimo.xiaomi.com/install | bash Windows : npm install -g @mimo-ai/cli After installation, run mimo in the terminal. The UI is optimized for iTerm or VS Code terminals on macOS.

Configuration commands are accessed via /, and all settings are fully localized in Chinese.

Model Support

MiMo Code ships with a limited‑time free channel for the MiMo‑V2.5 model and can also connect to DeepSeek, Kimi, GLM, or any third‑party token‑plan API.

Open‑Source Outlook

The project is released under the permissive MIT license, allowing individual developers to modify and redistribute, enterprises to integrate without licensing constraints, and the community to build vertical‑domain assistants.

Original Source

Signed-in readers can open the original source through BestHub's protected redirect.

Sign in to view source
Republication Notice

This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactadmin@besthub.devand we will review it promptly.

Open sourceAI coding assistantpersistent memorybenchmark comparisonCompose mode
Xiaomi Tech
Written by

Xiaomi Tech

Chat about technology with Xiaomi and change life together.

0 followers
Reader feedback

How this landed with the community

Sign in to like

Rate this article

Was this worth your time?

Sign in to rate
Discussion

0 Comments

Thoughtful readers leave field notes, pushback, and hard-won operational detail here.