Portal track
Official resources first
Open the Xiaomi MiMo blog, GitHub, Hugging Face, Studio, and API entries without hunting through fragmented launch materials.
Models, benchmarks, Studio, API, and source links
Xiaomi MiMo public models, benchmarks, Studio, and API in one clean portal.
This homepage organizes the public Xiaomi MiMo footprint the way an official resource center would feel: direct links first, model context second, and benchmark detail close to the source.
Portal track
Open the Xiaomi MiMo blog, GitHub, Hugging Face, Studio, and API entries without hunting through fragmented launch materials.
Portal track
See how MiMo, MiMo-7B, and MiMo-V2-Flash relate so the public model line reads like one family instead of disconnected repos.
Portal track
Keep reasoning, code, and agent results close to the official source tables so the portal stays useful and credible.
Models
A Mixture-of-Experts model positioned around fast reasoning, coding, long context, and agentic workflows.
The base MiMo-V2-Flash variant used as the public foundation for benchmark comparisons and downstream evaluation.
The reasoning-focused 7B public line from the earlier MiMo release, positioned around math and code performance.
Benchmarks
MiMo-V2-Flash is publicly framed as a reasoning-forward model family, with strong results on math and knowledge-heavy tasks.
MiMo-V2-Flash
84.9
MMLU-Pro
MiMo-V2-Flash
83.7
GPQA-Diamond
MiMo-V2-Flash
94.1
AIME 2025
Public benchmark tables emphasize coding, SWE-Bench, and terminal-style agent tasks as part of the MiMo-V2-Flash story.
MiMo-V2-Flash
80.6
LiveCodeBench-v6
MiMo-V2-Flash
73.4
SWE-Bench Verified
MiMo-V2-Flash
38.5
Terminal-Bench 2.0
Docs and explainers
Mar 20, 2026
A practical reading layer for MiMo-V2-Flash public benchmark tables across reasoning, code, and agent tasks.
Mar 20, 2026
The fastest route to Xiaomi MiMo public links: official blog, repositories, model hub, Studio, and API Platform.
Mar 20, 2026
A concise portal-style overview of MiMo-V2-Flash, Xiaomi's public reasoning, coding, and agent-focused model release.