Xiaomi Releases Open-Source AI Coding Assistant MiMo Code, Says It Outperforms Claude Code on Benchmarks

aiai-productsai-agents-coding 5 posts · 5 accounts

Xiaomi released MiMo Code, an open-source AI coding assistant for the terminal that the company said outperforms Claude Code on agentic coding and software-engineering benchmarks. The tool comes with MiMo V2.5, a multimodal model available free for a limited time with a 1 million-token context window aimed at larger coding projects.

MiMo Code is built to retain context across sessions through persistent memory and SQLite full-text search, with build, plan and compose modes and a MEMORY.md file for saved context. Xiaomi said the MIT-licensed software is designed to accumulate knowledge across large codebases, is compatible with Claude Code workflows and supports providers such as Anthropic, OpenAI, DeepSeek, Kimi and GLM.

From the sources (5 posts)

@wesroth

Xiaomi’s MiMo team released MiMo Code V0.1, an open-source AI coding agent for the terminal. The tool comes with MiMo V2.5, a multimodal model available free for a limited time, with a million-token context window for large coding projects

@0xchux

Xiaomi mimo code v 0.1.0 is live. Get started : 👇👇👇 curl -fsSL | bash I love the latest Xiaomi model - Mimo v2.5, it's my best design model and also good at planning and structuring tasks. Chinese models are too

@xiaomimimo

🚀 MiMo Code V0.1 is now live and open-source！ More than an AI coding assistant in your terminal — it's the smartest coding partner you'll ever work with. Comes with MiMo V2.5, a multimodal model available free for a limited time, featurin

@techmeme

Xiaomi releases MiMo Code V0.1.0, an open-source AI coding assistant that it says outperforms Claude Code on agentic coding and software engineering benchmarks (@carlfranzen / VentureBeat) (Visit Techmeme dot com for the link and full cont

@chenzeling4

Xiaomi open-sourced a coding agent that remembers what you were working on across sessions. MiMoCode runs in your terminal, uses SQLite full-text search for persistent memory, and has a free channel — no API keys, no setup. Three modes (bui

Market impact

22 small

NVDA MSFT GOOGL

Why it matters: An open-source, MIT-licensed coding agent from Xiaomi self-claiming to beat Claude Code is a marginal negative for the closed-lab coding-assistant margin thesis (Anthropic, proxied via MSFT/GOOGL) and AI-capex return case (NVDA), but it is a self-promotional launch with an unverified benchmark claim from a non-frontier lab and a tiny cluster, so it is not a DeepSeek-scale repricing.

At publish time: Headline states an open-source coding assistant claiming to outperform Claude Code on agentic coding benchmarks — a competitive-threat hook, but it is a company's own launch/PR claim, which the rules say to discount.

model confidence: medium

What happened next

1–2 weeks later

fizzled moved: flat

Xiaomi MiMo Code left no trace; follow-ups show Claude Code strengthening (value +27%, security fixes, new sync) plus routine open-source coding releases, no repricing.

based on follow-up coverage

Follow-up coverage

No follow-up coverage matched this story.

South Korean ETF Assets Hit Record $339 Billion as Kospi 200 Bearish Bets Reach 5-Year High Anthropic Releases Claude Fable 5, Public Version of Mythos, at $10/$50