Back

Xiaomi

MiMo-V2.5

ReasoningOpen weightsModels.dev
Intelligence
21.8
Rank 222 of 747 measured models
Coding
56.8
Rank 98 of 298 measured models
Output speed
36.1 t/s
Rank 379 of 405 measured models
Price/1M
$0.175
Rank 146 of 664 measured models
Context window
1M
Rank 84 of 1207 measured models
Cost per task
$0.0191
Rank 21 of 196 measured models

MiMo-V2.5 scores 21.8 on the Artificial Analysis Intelligence Index, ranking 222nd of 747 measured models. The closest models on this index are o3-pro (21.9), G9v3-39A5B (21.8), and KAT Coder Pro V2 (21.7). At $0.175 per million tokens (3:1 input/output blended price), it is cheaper than o3-pro ($35) and KAT Coder Pro V2 ($0.525).

Artificial Analysis Indices
Intelligence21.8
Rank 222 of 747 measured models
Coding56.8
Rank 98 of 298 measured models
Agentic15.8
Rank 109 of 190 measured models
Capability indexes
Performance in specific industries, measured by Artificial Analysis
Finance & Accounting22.2
Rank 151 of 222 measured models
Strategy & Ops18.9
Rank 151 of 222 measured models
Legal24.4
Rank 152 of 222 measured models
Engineering24.6
Rank 144 of 221 measured models
Economics33.9
Rank 139 of 229 measured models
Standard Benchmarks
GPQA Diamond84.9%
Rank 138 of 620 measured models
HLE27.2%
Rank 180 of 667 measured models
SciCode43.9%
Rank 152 of 234 measured models
IFBench67.1%
Rank 90 of 450 measured models
AA-LCR73.0%
Rank 185 of 578 measured models
TerminalBench Hard41.7%
Rank 50 of 432 measured models
Terminal-Bench 2.163.7%
Rank 93 of 245 measured models
Terminal-Bench v40.0%
Rank 151 of 221 measured models
Additional benchmarks8
Design Arena · 3d50.8%
Design Arena · Asciiart45.0%
Design Arena · Codecategories53.5%
Design Arena · Dataviz53.6%
Design Arena · Gamedev52.6%
Design Arena · Svg50.0%
Design Arena · Uicomponent53.7%
Design Arena · Website53.6%
Models at a similar level
The closest models on the Intelligence Index, one configuration per family.

Preferences and analytics

Rybbit: visits, performance, clicks, copies, forms and session replay (masked inputs), with your permission. Refusing does not limit the site.

Privacy

Allow navigation analytics?

With your permission, Rybbit measures visits, performance and interactions (clicks, copies, forms), including session replay, to improve BenchSift. Input values are masked in replays. Refusing does not limit the site.

Your choice is stored for 180 days. Change it at any time using “Privacy preferences” in the footer. Withdrawal reloads the page.

No choice stored: Rybbit is disabled.

Privacy