Models & Releases

Xiaomi Drops Stealth Trillion-Parameter Hunter Alpha – The Mystery Model Everyone's Buzzing About!

The mystery 'Hunter Alpha' model that topped OpenRouter charts was Xiaomi's trillion-parameter agent AI in disguise.

Deep Dive

Xiaomi's AI division MiMo pulled off one of the most successful stealth launches in recent AI history with its trillion-parameter MiMo-V2-Pro model. For over a week, the model operated anonymously on OpenRouter as 'Hunter Alpha,' racking up over one trillion tokens processed during testing while developers speculated it was DeepSeek's upcoming V4 model. The reveal caused Xiaomi's Hong Kong shares to surge 5.8%, demonstrating the market impact of this surprise entry into the high-stakes AI race.

MiMo-V2-Pro represents a strategic shift toward agentic AI rather than conversational chatbots. Designed specifically for complex, multi-step workflows requiring planning and tool use, the model ranks third globally on OpenClaw's PinchBench and ClawEval benchmarks—trailing only Claude Sonnet 4.6 and Claude Opus 4.6. Xiaomi's internal engineers report coding performance approaching Claude Opus 4.6 levels, while pricing comes in approximately 67% cheaper than Sonnet 4.6 for standard contexts.

The release includes two companion models: MiMo-V2-Omni (previously 'Healer Alpha'), a full multimodal agent model, and MiMo-V2-TTS for natural voice output. All three are accessible through Xiaomi's MiMo API platform, with MiMo-V2-Pro remaining free on OpenRouter for a limited time. This positions Xiaomi as a serious contender in the agent AI space, particularly for developers building with frameworks like OpenClaw, OpenCode, and KiloCode who need cost-effective, high-performance reasoning capabilities.

Key Points
  • MiMo-V2-Pro processed over 1 trillion tokens during anonymous testing as 'Hunter Alpha' on OpenRouter
  • Ranks 3rd globally on agent benchmarks (OpenClaw's PinchBench/ClawEval) behind Claude Sonnet 4.6 and Opus 4.6
  • Priced 67% cheaper than Claude Sonnet 4.6 with 1M token context and agent-focused architecture

Why It Matters

Delivers Claude-level agent performance at significantly lower cost, accelerating practical AI automation for developers and enterprises.