Xiaomi open-sources MiMo Code AI coding agent, claims it outperforms Claude Code on complex 200-step software tasks
Xiaomi has unveiled MiMo Code, a new open-source AI coding agent designed to handle long and complex software engineering workflows. The company claims the tool can outperform Anthropic’s Claude Code on ultra-long coding tasks involving up to 200 sequential steps, a scenario that often challenges AI coding assistants. Released under the MIT license, MiMo Code is built to help developers manage large-scale projects while maintaining context and consistency across extended development sessions.
Xiaomi makes MiMo Code available to developers
Xiaomi announced the release of MiMo Code on June 10 through its official @XiaomiMiMo account on X, describing the tool as “more than an AI coding assistant in your terminal.” The company positioned MiMo Code as a coding partner designed to help developers handle complex software engineering workflows.
The coding agent is now available on GitHub under the MIT license, allowing developers to access, modify and distribute the software. Xiaomi said MiMo Code can be installed on macOS and Linux using a single terminal command, while Windows users can install it via npm using the @mimo-ai/cli package. The broad platform support is aimed at making the coding agent accessible to developers across different operating systems.
Xiaomi targets a key weakness in AI coding tools
AI coding assistants have rapidly evolved from simple code-completion tools into systems capable of generating applications, debugging software and automating repetitive development tasks. However, many of these tools still struggle when projects become larger and require hundreds of interconnected actions.
According to Xiaomi, MiMo Code has been developed specifically to address this limitation. The company says the coding agent is designed to maintain context across long-running projects, helping developers work through complex software engineering tasks without repeatedly losing track of earlier decisions or requirements.
Rather than functioning as a conventional chatbot that responds to isolated prompts, MiMo Code is intended to operate as a software engineering agent capable of handling extended workflows. Xiaomi believes this capability will become increasingly important as AI tools take on larger responsibilities throughout the software development lifecycle.
Xiaomi claims stronger performance on long-horizon coding tasks
The headline claim surrounding MiMo Code centres on its ability to tackle what Xiaomi describes as long-horizon software engineering tasks.
According to the company, the coding agent demonstrated stronger performance than Claude Code during evaluations involving workflows that required up to 200 consecutive actions. These types of tasks are considered particularly challenging because AI systems must maintain awareness of project goals, previous decisions and evolving requirements over extended periods.
Xiaomi says MiMo Code has been designed to preserve project context throughout these lengthy workflows rather than treating each interaction as a separate task. This allows the system to build on previous work while continuing toward a larger objective.
More than a code-generation tool
Xiaomi describes MiMo Code as a complete coding agent rather than a traditional AI assistant.
The platform is designed to support multiple stages of software development, including project planning, implementation, testing and reporting. Instead of simply responding to prompts, the system can work through structured workflows that involve multiple interconnected tasks.
According to Xiaomi, MiMo Code operates through a terminal-based interface and is built on the open-source OpenCode project. The company says this approach provides developers with greater flexibility and control over how the system is integrated into their workflows.
Another notable feature is support for multiple AI models. While MiMo Code is designed around Xiaomi’s own MiMo models, developers are not limited to a single provider. The platform can also connect to third-party AI services, giving users more flexibility when choosing the models that best suit their requirements.
An open-source strategy could help drive adoption
One of the most significant aspects of Xiaomi’s announcement is the decision to release MiMo Code as open-source software under the MIT license.
The open-source approach allows developers and organisations to inspect, modify and distribute the software while adapting it to their own requirements. At a time when many AI coding tools operate behind proprietary platforms and subscription models, Xiaomi is taking a markedly different route.
The company says MiMo Code is intended to serve as a foundation for developers building their own workflows and agent-based software engineering systems. Open access could also encourage community contributions and faster iteration as developers experiment with new features and integrations.