MiMo V2 Pro
Editorial notes
Lanzado 18 marzo 2026 (codename Hunter Alpha desde 11 marzo en OpenRouter, donde fue confundido con DeepSeek V4). Flagship de Xiaomi. MoE 1T+ totales / 42B activos. Hybrid Attention 7:1 SWA/GA + Multi-Token Prediction. Liderado por Luo Fuli (ex-DeepSeek). Optimizado para agentic tasks. Ranking #2 modelos chinos en AA Intelligence Index. 5-8x mas barato que Claude Opus 4.6 ($1/$3 per MTok). Hallucination 30% (vs V2-Flash 48%). Reduce el 5.8% subio el stock de Xiaomi tras el reveal.
Spec sheet
- Company
- Xiaomi
- Country
- CN
- Type
- reasoning
- Release
- 2026-03
- Context
- 1.0M tokens
- Params total
- 1.0T
- Params active (MoE)
- 42B
- License
- proprietary
- Pricing (openrouter)
- $1/$3/M
- Slug
- mimo-v2-pro
Benchmarks (4)
Reasoning 2
Cite this model
BibTeX · APA
BibTeX
@misc{frontier-mimo-v2-pro,
title = {MiMo V2 Pro},
author = {{Xiaomi}},
year = {2026},
note = {Frontier Benchmarks AI atlas. Accessed 2026-05-08},
url = {https://frontierbenchmarks.com/models/mimo-v2-pro}
} APA
Xiaomi (2026). MiMo V2 Pro [Large language model]. Frontier Benchmarks AI. Retrieved 2026-05-08, from https://frontierbenchmarks.com/models/mimo-v2-pro
Citation reflects the atlas page, not the original model paper. For the paper, see the "Resources" section above.