‹ PublicAI IndexGoogle
The LLM benchmark aggregator.
Gemini 3.1 Pro
Strongest in Science (#1 of 122), weakest in Agents (#197 of 298). Above par in 16 of 23 scopes. Among the models it meets almost everywhere, it finishes behind Claude Opus 5.5 and Claude Fable 5.1 and ahead of DeepSeek V4.1 Flash and GPT-5.6 Terra.
§ 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.
Core abilities60.6−4.7#11/1991/4
Instruction following69.7−4.1#3/621/1
Language59.6−7#12/621/1
Human preference65−3.8#16/3541/1
Human preference65−3.8#16/3541/1
Knowledge56.4−6.1#28/1972/3
Academic knowledge62.2−1.7#4/1231/1
Document parsing55.7−7.6#32/831/1
Reasoning56.8−8.6#32/1954/5
Science63.5leads#1/1221/1
Reasoning62.6−6.4#12/1472/3
Mathematics51−12.8#80/1522/2
Professional53.8−8.6#46/1941/2
Medical59.7−6.8#10/1491/1
Biology research36.1−34.6#29/311/1
Legal53.8−12.1#57/1731/1
Finance54.2−8.2#61/1621/1
Coding45.5−20.7#122/1792/6
Agentic coding45.3−21.6#117/1732/5
Safety49.6−11.8#187/3291/3
Factual grounding48.8−21.9#61/1041/1
Agents46.2−20.2#197/2982/6
Knowledge work43.9−26.8#131/2082/2
§ 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 Gemini 3.1 Pro, left for the other.
§ 3 · Sources
Where the numbers come from
9 publications, 38 figures. Every one links to the page it was read from.
LMArena Text 1487
GDPval-AA 799
AA-Briefcase 457
ARC-AGI-2 77.1%
LiveBench 1.0 77LiveBench 1.0 · Reasoning 84LiveBench 1.0 · Coding 76.5LiveBench 1.0 · Agentic Coding 44.1LiveBench 1.0 · Mathematics 91LiveBench 1.0 · Data Analysis 78.5LiveBench 1.0 · Language 85.4LiveBench 1.0 · Instruction Following 79.1
SimpleBench 79.6%
Vals · Legal Research Bench 20.67%Vals · CaseLaw 64.84%Vals · LegalBench 87.4%Vals · Finance Agent 42.98%Vals · CorpFin 64.49%Vals · TaxEval 72.88%Vals · MortgageTax 69.4%Vals · MedQA 96.37%Vals · MedCode 59.06%Vals · MedScribe 76.11%Vals · BioMysteryBench 60%Vals · SWE-bench Verified 78.8% (Mini-SWE-agent)Vals · Vibe Code Bench 32.03% (OpenHands)Vals · Code Migration 17.31%Vals · GPQA Diamond 95.45%Vals · MMLU Pro 90.99%Vals · AIME 98.13%Vals · ProofBench 26%
ParseBench · Overall 67.32%ParseBench · Tables 91.13%ParseBench · Charts 40.43%ParseBench · Content faithfulness 89.26%ParseBench · Semantic formatting 44.91%ParseBench · Visual grounding 70.85%
Vectara · Factual consistency 89.6%
Badge
[](https://publicai.io/model-index/m/gemini-3-1-pro)