On June 8, 2026, Xiaomi’s MiMo-V2.5-Pro-UltraSpeed crossed 1,000 tokens per second on standard, rentable cloud GPUs, a milestone never reached before at the trillion-parameter scale. The current flagship, MiMo-V2.5-Pro, released on April 22, 2026, operates on a 1.02-trillion-parameter Mixture-of-Experts architecture, supports a 1-million-token context window, and processes text, image, audio, and video natively under an MIT open-source licence. Pricing is $1.00 per million input tokens and $3.00 per million output tokens, compared to Claude Opus 4.6’s $5.00 input and $25.00 output. MiMo-V2.5-Pro scores 57.2 on SWE-bench Pro versus Claude Opus 4.6’s 53.4, using 40–60% fewer tokens for comparable agentic tasks. Before official naming, it was deployed anonymously as “Hunter Alpha” on OpenRouter and topped daily usage rankings. The UltraSpeed mode achieves 1,000+ tokens per second (peak ~1,200), while GPT-5.5 runs at ~68, Claude Opus 4.6 at ~71, Claude Haiku at ~98, and Gemini Flash at ~192 tokens per second.