‹ PublicAI IndexGoogle
The LLM benchmark aggregator.
Gemini 4 Argon
Strongest in Human preference (#1 of 350), weakest in Safety (#225 of 331). Above par in 22 of 25 scopes. Among the models it meets almost everywhere, it finishes behind GPT-6 Astra and Claude Opus 5 and ahead of Claude Sonnet 5.5 and Muse Wonder Full.
§ 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.
Human preference68.6leads#1/3501/1
Human preference68.6leads#1/3501/1
Professional63.1leads#1/1882/2
Legal65.9leads#1/1641/1
Medical64−2.5#2/1491/1
Finance61.4−1.1#3/1611/1
Cybersecurity61.7−2.3#3/461/1
IT operations72.1−1.7#3/311/1
Biology research61.1−7.1#6/281/1
Document design49.7−17.6#30/481/1
Core abilities64.4−0.8#4/2252/4
Data analysis66.1leads#1/481/1
General intelligence66.5−2.3#8/2142/3
Reasoning62.1−2.8#7/1903/5
Reasoning64−2.7#7/1442/3
Mathematics61.5−2.6#9/1501/2
Coding62−4.4#7/1762/6
Agentic coding62.6−4.4#7/1692/5
Agents63−2.9#9/2893/6
Long-horizon work67.6−6.4#6/481/1
Computer use61.5−6.2#9/811/2
Web research58.7−3.9#10/491/2
Knowledge work65.5−6.3#19/1982/2
Safety48.3−13.1#225/3311/3
Factual grounding44.7−26.1#78/1021/1
§ 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 4 Argon, left for the other.
§ 3 · Sources
Where the numbers come from
8 publications, 26 figures. Every one links to the page it was read from.
LMArena Text 1524
Artificial Analysis Intelligence Index 53
GDPval-AA 1623
AA-Briefcase 1486
ARC-AGI-2 91.3%
LiveBench 84.1LiveBench · Agentic 69.9LiveBench · Reasoning 95.1LiveBench · Coding 89.7LiveBench · Browser Use 92LiveBench · Data Analysis 85.2LiveBench · Research 83.1LiveBench · Design 73.7
Vals · Legal Research Bench 54.81%Vals · LegalBench 88.3%Vals · Harvey Legal Agent Benchmark 27.08%Vals · Finance Agent 65.4%Vals · MedCode 58.8%Vals · MedScribe 87.43%Vals · BioMysteryBench 76.3%Vals · CyberBench 77.86%Vals · SRE Bench 44.27%Vals · Vibe Code Bench 91.91% (OpenHands)Vals · Code Migration 68.17%Vals · ProofBench 99%
Vectara · Factual consistency 87.9%
Badge
[](https://publicai.io/model-index/m/gemini-4-argon)