System Card← leaderboard
Google DeepMind

Gemini Argon

#4of 28 · general

Gemini 4 Argon

$2 / $10per 1M in / out
2026-09-30released
50100Coding 73 · #6 of 28Agents 81 · #4 of 29Cyber 44 · #2 of 24Engineering 72 · #7 of 26Math 76 · #6 of 26Health 70 · #5 of 24Legal 88 · #1 of 27Professional 79 · #4 of 28Writing 74 · #3 of 21 · one sourceInstruct 91 · #1 of 25Hallucination 53 · #9 of 26Multimodal 80 · #3 of 24
Coding73#6
Agents81#4
Cyber44#2
Engineering72#7
Math76#6
Health70#5
Legal88#1
Professional79#4
Writing74#3
Instruct91#1
Hallucination53#9
Multimodal80#3

Effort

Gemini 4 Argon
score %49.55050.551$ / task · log scaleMedHigh

Measured

≈ imputed · amber dot = lab-self-reported among the sources
GPQA Diamond96.4%#1Epoch AI
HLE57.1%#3Artificial Analysis
SciCode61.8%#3Artificial AnalysisEpoch AI
CritPt27.1%#7Artificial AnalysisEpoch AI
Terminal-Bench57.7%#17Vals.ai
APEX Agents82.4%#1Epoch AI
GDPval56.2%#12Artificial Analysis
Arena1525#1Arena
AA Index52.6#7Artificial Analysis
Vals Index68.9#1Vals.ai
Epoch Capabilities Index166.4#5Epoch AI

Neighbors

Claude Fable→ looser instructions · pricierClaude Opus→ looser instructions · pricierGPT Astra→ more factual · pricierMuse Spark→ weaker agent · cheaperMuse Wonder→ weaker multimodal · pricier