The weeks-long mystery surrounding "Hunter Alpha" — an anonymous AI model that appeared on OpenRouter and quietly climbed to the top of multiple agent benchmarks — is over. On March 19, Xiaomi officially confirmed that Hunter Alpha is an early test build of MiMo-V2-Pro, the company's flagship foundation model designed for the agentic AI era.
Xiaomi CEO Lei Jun made the announcement alongside the release of two companion models: MiMo-V2-Omni, a multimodal agent model, and MiMo-V2-TTS, a speech synthesis model.
A Trillion Parameters, 42 Billion Active
MiMo-V2-Pro houses over 1 trillion total parameters but uses a mixture-of-experts architecture where only 42 billion parameters are active during any single forward pass. That makes it roughly three times the size of its predecessor, MiMo-V2-Flash, while remaining efficient enough for production deployment. The model supports a context window of up to 1 million tokens.
The MiMo team is led by Luo Fuli, a former DeepSeek researcher, and described MiMo-V2-Pro as an "agent brain" — built to handle complex, multi-step workflows with fewer human prompts and interventions than a standard chatbot.
Benchmarks That Turned Heads
What made Hunter Alpha impossible to ignore was its performance. On ClawEval, a benchmark for agentic scaffolds, the model scored 61.5 — significantly outpacing GPT-5.2 at 50.0 and approaching Claude Opus 4.6's 66.3. On PinchBench, it achieved an average score of 81.0, placing third globally behind both Claude 4.6 variants.
Among the 160 models in its price tier (under $0.15 per million tokens) on OpenRouter, MiMo-V2-Pro ranks first with a score of 49 on the Artificial Analysis Intelligence Index, far exceeding the category median of 13.
The Cost Advantage
Pricing is where MiMo-V2-Pro makes its strongest case. The model runs at $1 per million input tokens and $3 per million output tokens for contexts up to 256K, scaling to $2/$6 for contexts between 256K and 1 million tokens. By comparison, running the same ClawEval index cost $348 with MiMo-V2-Pro versus $2,304 for GPT-5.2 and $2,486 for Claude Opus 4.6.
The multimodal MiMo-V2-Omni model is even cheaper at $0.40 per million input tokens and $2.00 per million output tokens.
What This Means for the AI Landscape
Xiaomi's entry into the frontier model race signals that competition in agentic AI is no longer a two-horse race between U.S. labs. With performance approaching Western frontier models at a fraction of the cost, MiMo-V2-Pro could accelerate enterprise AI adoption — particularly in Asia, where Xiaomi already has massive distribution through its consumer hardware ecosystem.
The reveal also highlights a growing trend: Chinese AI labs are closing the gap with U.S. competitors faster than many analysts predicted, and they are doing it with aggressive cost optimization that could reshape pricing expectations across the entire industry.
