‹ PublicAI Index
The LLM benchmark aggregator.
MiMo V2.5
Xiaomi · 311B · open weights
Strongest in Science (#64 of 122), weakest in Professional (#151 of 194). Above par in 5 of 15 scopes. Among the models it meets almost everywhere, it finishes behind Claude Fable 5 and Claude Fable 5.1 and ahead of Gemini 3.1 Flash-Lite and DeepSeek V3.
§ 1 · Profile
What it is good at
Bars run from 50 — the average of the models each source lists — so right of the line is above par. The middle column is the gap to whoever leads that scope.
Human preference59.9−8.9#89/3541/1
Human preference59.9−8.9#89/3541/1
Knowledge50.9−11.6#102/1971/3
Academic knowledge52.1−11.8#71/1231/1
Coding45.9−20.3#116/1791/6
Agentic coding45.6−21.3#112/1731/5
Reasoning45.7−19.7#132/1951/5
Science53.8−9.7#64/1221/1
Mathematics38.3−25.5#125/1521/2
Agents49.7−16.7#132/2982/6
Knowledge work49.6−21.1#98/2082/2
Professional45.2−17.2#151/1941/2
Finance48.1−14.3#109/1621/1
Medical40.6−25.9#131/1491/1
Legal42.8−23.1#141/1731/1
§ 2 · Head to head
What it beats, and what beats it
The same models turn up scope after scope. Counted once: where both were placed, who finished higher. Bars run right for MiMo V2.5, left for the other.
§ 3 · Sources
Where the numbers come from
4 publications, 17 figures. Every one links to the page it was read from.
LMArena Text 1434
GDPval-AA 1005
AA-Briefcase 754
Vals · Legal Research Bench 9.13%Vals · LegalBench 78.89%Vals · Finance Agent 36.73%Vals · CorpFin 59.91%Vals · TaxEval 71.83%Vals · MortgageTax 59.26%Vals · MedCode 31.89%Vals · MedScribe 72.15%Vals · SWE-bench Verified 71% (Mini-SWE-agent)Vals · Vibe Code Bench 42.17% (OpenHands)Vals · Code Migration 14.23%Vals · GPQA Diamond 81.57%Vals · MMLU Pro 82.93%Vals · ProofBench 16%
Badge
[](https://publicai.io/model-index/m/mimo-v2-5)