Coverage (30d)
0vs6
This Week
0vs2
Evidence
1 articlesRelationships
1Timeline
Gemini 3 Pro2026-04-16
Achieved top score on METR time horizon benchmark, handling 90-minute software tasks
Gemini 3 Pro2026-02-20
Achieved state-of-the-art status on most benchmarks according to preliminary evaluations
Ecosystem
CUDA Agent
competes withGemini 3 Pro1 src
competes withtorch.compile1 src
usesCUDA1 src
usesreinforcement learning1 src
competes withClaude Opus 4.61 src
Gemini 3 Pro
usesMMLU1 src
competes withGPT-4 Turbo1 src
competes withClaude 31 src
Benchmarks
mmlu pro
CUDA Agent—
Gemini 3 Pro90.1
arena elo
CUDA Agent—
Gemini 3 Pro1485
swe bench verified
CUDA Agent—
Gemini 3 Pro80.6