Chinese electronics manufacturer Xiaomi has released MiMo Code V0.1.0, an open-source AI coding assistant that the company claims outperforms Anthropic's Claude Code on extended programming tasks requiring 200 or more steps. The terminal-based tool was announced June 10, 2026, and is available under an MIT license on GitHub.

According to Xiaomi's internal testing with 576 developers across 474 repositories, MiMo Code matched Claude Code's performance on tasks under 200 steps but achieved a win rate above 65% on longer tasks. The company reported benchmark scores of 82% on SWE-bench Verified versus Claude Code's 79%, and 62% versus 55% on SWE-bench Pro. However, these figures have not been independently verified, and Xiaomi did not compare against other major competitors like OpenAI's Codex or Google's Gemini CLI.

The system addresses a common problem in AI coding agents: memory degradation during extended sessions. MiMo Code employs a cross-session memory architecture using SQLite full-text search, maintaining project memory through persistent files, session checkpoints, and task progress logs. The tool includes voice control capabilities and automatically imports configurations from Claude Code for easy migration.

MiMo Code ships with free access to Xiaomi's MiMo-V2.5 model, which features a one-million-token context window. The company positions this as a limited-time offering, with pricing starting at $0.40 per million input tokens for the base model. The tool supports third-party model backends and requires zero initial configuration.

Separately, NanoClaw, an enterprise-focused variant of the open-source OpenClaw agent, announced a security partnership with software supply chain company JFrog. The integration aims to prevent AI agents from downloading malicious code by routing all package requests through JFrog's vetted registries. When agents attempt to install potentially harmful dependencies, the system blocks the request and guides them to approved alternatives.

The NanoClaw-JFrog partnership addresses growing concerns about autonomous agents installing packages without human oversight. The integration is free for open-source users and available for enterprise customers through existing JFrog commercial licenses. This follows previous NanoClaw security initiatives including permission dialogs and Docker container isolation.

Both developments reflect the rapidly evolving landscape of AI coding tools, where Chinese companies are increasingly competing with established U.S. providers through aggressive pricing and open-source releases, while security concerns around autonomous agent behavior drive new protective measures.