Back

Google

Gemma 4 31B

ReasoningOpen weightsHugging Face
Intelligence
14.7
Rank 314 of 744 measured models
Coding
43.4
Rank 148 of 298 measured models
Output speed
35.5 t/s
Rank 378 of 403 measured models
Context window
256K
Rank 518 of 1199 measured models
Cost per task
$0
Rank 1 of 193 measured models

Gemma 4 31B (Reasoning) scores 14.7 on the Artificial Analysis Intelligence Index, ranking 314th of 744 measured models. The closest models on this index are Grok 3 mini Reasoning (High) (14.6), DeepSeek V3.2 Speciale (14.5), and Gemini 2.5 Pro Preview (May' 25) (14.5). With an output speed of 35.5 t/s, it is slower than Grok 3 mini Reasoning (High) (65.3 t/s).

Artificial Analysis Indices
Reasoning
Intelligence14.7
Rank 314 of 744 measured models
Coding43.4
Rank 148 of 298 measured models
Agentic4.2
Rank 133 of 189 measured models
Reasoning levels
Each point is a reasoning level of this model: more reasoning often raises the score but costs more. Select a point to open that level.
Intelligence Indexmost efficient ↗Output speed
Capability indexes
Performance in specific industries, measured by Artificial Analysis
Finance & Accounting13.6
Rank 173 of 218 measured models
Strategy & Ops9.3
Rank 173 of 218 measured models
Legal13.9
Rank 179 of 218 measured models
Engineering16.7
Rank 175 of 217 measured models
Economics20.6
Rank 181 of 225 measured models
Decision Index
Gemma 4 31B (Reasoning) · Artificial Analysis · measured via DeepInfra
Decision Index55.5Rank 14 of 26 measured models
AA-Omniscience-Decision · accuracy47.6
AA-Omniscience-Decision · non-hallucination4.2
HLE-MultiChoice25.7
When2Call77.6
WildGuardTest92.8
Cost per 1M tasks$624.44
Average time per task14.4 s
Decision Index leaderboard
Standard Benchmarks
GPQA Diamond85.7%
Rank 127 of 619 measured models
HLE23.6%
Rank 195 of 664 measured models
SciCode45.5%
Rank 140 of 231 measured models
IFBench75.6%
Rank 30 of 450 measured models
AA-LCR69.7%
Rank 225 of 575 measured models
TerminalBench Hard36.4%
Rank 72 of 432 measured models
Terminal-Bench 2.143.4%
Rank 139 of 244 measured models
Terminal-Bench v40.0%
Rank 151 of 218 measured models
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