Xiaomi challenges ai giants with surprisingly cheap llm
Xiaomi, best known for its aggressively priced smartphones, is making a bold move into the artificial intelligence arena with MiMo-V2-Pro, a large language model that’s undercutting the likes of OpenAI and Anthropic on price while delivering surprisingly competitive performance. It's a development that could reshape the landscape of AI accessibility, particularly for developers and smaller businesses.
The numbers don't lie: mimo-v2-pro's performance
Initially launched in March, MiMo-V2-Pro didn't immediately set the AI world ablaze. But recent benchmarks tell a different story. Xiaomi claims its model now rivals, and in some instances surpasses, established players like Claude Sonnet 4.6, Google’s Gemini 3 Pro, and even early iterations of GPT-5.2. This isn't just marketing hype; the company asserts MiMo-V2-Pro delivered superior results in preliminary testing even before its current iteration, consistently outperforming Sonnet 4.6 across a range of scenarios.
The architecture itself is noteworthy: MiMo-V2-Pro boasts over a billion parameters, with 42 billion actively engaged, and supports up to 1 million tokens—a level of input capacity on par with some of the most advanced models currently available. But the real shocker is the cost.
A price point that's disrupting the market
While initially offered for free, MiMo-V2-Pro now operates on a token-based pricing model. Even so, the rates are astonishingly low. Users can pay as little as $1 per million input tokens with a 256,000 token context window, or $2 per million tokens with a full 1 million token context. Output tokens run between $3 and $6 per million—a significant reduction compared to Anthropic's Claude Sonnet 4.6 and Opus 4.6, which charge between $3 and $5 per million input tokens and a staggering $15 to $25 per million output tokens. Consider this: Xiaomi sells a smartphone for around €75, and the MiMo-V2-Pro model is powerful enough to even drive OpenClaw, a popular open-source AI game engine.
The move isn't just about price. Anthropic recently restricted Claude’s integration with third-party tools like OpenClaw due to excessive token consumption—a limitation Xiaomi is actively circumventing by leveraging MiMo-V2-Pro to power OpenClaw directly. While Claude can still connect, users seeking that functionality now face additional subscription costs. Xiaomi’s strategy, in contrast, is one of open access and developer empowerment.
Perhaps the most interesting element is Xiaomi’s willingness to buck trends. While other AI developers are focused on ever-increasing model size and complexity, Xiaomi has demonstrated that a powerful, cost-effective AI model can be built without breaking the bank. This could democratize access to advanced AI capabilities, allowing smaller companies and individual developers to compete on a more level playing field.
The company's foray into AI is a reminder that innovation isn't confined to Silicon Valley giants. Xiaomi, with its focus on affordability and accessibility, is quietly reshaping the landscape of AI, one remarkably cheap token at a time.