Back

Google

Gemini 2.5 Pro

ReasoningClosed weightsModels.dev
Intelligence
16.1
Rank 295 of 744 measured models
Coding
33.3
Rank 175 of 298 measured models
Output speed
110 t/s
Rank 157 of 403 measured models
Price/1M
$3.44
Rank 510 of 659 measured models
Context window
1M
Rank 84 of 1199 measured models
Cost per task
$0.33
Rank 88 of 193 measured models

Gemini 2.5 Pro scores 16.1 on the Artificial Analysis Intelligence Index, ranking 295th of 744 measured models. The closest models on this index are Claude 4 Opus (16.6), DeepSeek V3.2 (16.0), and MiMo-V2-Flash (16.0). At $3.44 per million tokens (3:1 input/output blended price), it is cheaper than Claude 4 Opus ($30), more expensive than DeepSeek V3.2 ($0.315).

Artificial Analysis Indices
Intelligence16.1
Rank 295 of 744 measured models
Coding33.3
Rank 175 of 298 measured models
Math87.7
Rank 40 of 270 measured models
Agentic1.6
Rank 143 of 189 measured models
Capability indexes
Performance in specific industries, measured by Artificial Analysis
Finance & Accounting17.3
Rank 162 of 218 measured models
Strategy & Ops12.6
Rank 166 of 218 measured models
Legal23.7
Rank 151 of 218 measured models
Engineering18.9
Rank 164 of 217 measured models
Economics24.5
Rank 167 of 225 measured models
Standard Benchmarks
MMLU Pro86.2%
Rank 18 of 349 measured models
GPQA Diamond84.4%
Rank 150 of 619 measured models
HLE22.5%
Rank 202 of 664 measured models
LiveCodeBench80.1%
Rank 28 of 343 measured models
SciCode46.3%
Rank 136 of 231 measured models
MATH-50096.7%
Rank 29 of 202 measured models
AIME 202488.7%
Rank 10 of 194 measured models
AIME 202587.7%
Rank 40 of 270 measured models
Additional benchmarks6
Design Arena · 3d50.6%
Design Arena · Codecategories57.5%
Design Arena · Dataviz68.2%
Design Arena · Gamedev54.2%
Design Arena · Uicomponent57.5%
Design Arena · Website58.4%
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