‹ PublicAI IndexGoogle
The LLM benchmark aggregator.
Gemini 3.7 Flash
Strongest in Instruction following (#2 of 62), weakest in Mathematics (#68 of 152). Above par in 20 of 22 scopes. Among the models it meets almost everywhere, it finishes behind Gemini 4 Argon and Claude Fable 5 and ahead of GPT-5.5 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 abilities61.2−4.1#9/1991/4
Instruction following71.1−2.7#2/621/1
Language59.8−6.8#10/621/1
Human preference65.2−3.6#13/3541/1
Human preference65.2−3.6#13/3541/1
Knowledge58−4.5#17/1972/3
Academic knowledge61.1−2.8#6/1231/1
Document parsing57.8−5.5#24/831/1
Reasoning57−8.4#31/1953/5
Science62.5−1#5/1221/1
Reasoning59−10#23/1471/3
Mathematics53.5−10.3#68/1522/2
Agents54.9−11.5#57/2982/6
Knowledge work57.8−12.9#55/2082/2
Professional53.1−9.3#60/1941/2
Medical59.1−7.4#15/1491/1
IT operations42.5−31.3#18/331/1
Finance57.8−4.6#19/1621/1
Cybersecurity30.4−33.5#42/471/1
Legal55.4−10.5#44/1731/1
Coding52.7−13.5#63/1793/6
Agentic coding52.8−14.1#59/1733/5
§ 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 3.7 Flash, left for the other.
§ 3 · Sources
Where the numbers come from
8 publications, 34 figures. Every one links to the page it was read from.
LMArena Text 1489
GDPval-AA 1404
AA-Briefcase 1113
Terminal-Bench 12.4% (mini-SWE-agent)
ARC-AGI-2 84.6%
LiveBench 1.0 78.8LiveBench 1.0 · Reasoning 87.8LiveBench 1.0 · Coding 78.9LiveBench 1.0 · Agentic Coding 58.3LiveBench 1.0 · Mathematics 93.5LiveBench 1.0 · Data Analysis 68LiveBench 1.0 · Language 85.5LiveBench 1.0 · Instruction Following 79.9
Vals · Legal Research Bench 34.62%Vals · LegalBench 87.26%Vals · Finance Agent 59.04%Vals · TaxEval 74.73%Vals · MortgageTax 66.65%Vals · MedCode 53.39%Vals · MedScribe 83.94%Vals · CyberBench 43.75%Vals · SRE Bench 4.58%Vals · SWE-bench Verified 80.8% (Mini-SWE-agent)Vals · Vibe Code Bench 70.39% (OpenHands)Vals · Code Migration 34.8%Vals · GPQA Diamond 93.94%Vals · MMLU Pro 90.12%Vals · ProofBench 58%
ParseBench · Overall 70.37%ParseBench · Tables 90.21%ParseBench · Charts 32.07%ParseBench · Content faithfulness 88.21%ParseBench · Semantic formatting 74.98%ParseBench · Visual grounding 66.4%
Badge
[](https://publicai.io/model-index/m/gemini-3-7-flash)