Benchmark

Google — reported benchmarks

Google · source page ↗ · last checked Nov 9, 2026, 12:04 PM

Reported benchmarks · Gemini 3.1 Pro

captured Nov 9, 2026, 12:04 PM
BenchmarkScore
GPQA Diamond94.3% · No tools, Thinking (High)
SWE-Bench Verified80.6% · Single attempt, Thinking (High)
SWE-Bench Pro (Public)54.2% · Single attempt, Thinking (High)
Humanity's Last Exam44.4% · No tools, Thinking (High)
Humanity's Last Exam51.4% · Search (blocklist) + Code, Thinking (High)
ARC-AGI-277.1% · Abstract reasoning puzzles, Thinking (High)
Terminal-Bench 2.068.5% · Agentic terminal coding, Terminus-2 harness, Thinking (High)
LiveCodeBench Pro2887 Elo · Competitive coding, Thinking (High)

Vendor-reported via automated web search — not independently verified. See the cited matrix on /models.

In the news · Google

Importance-filtered press coverage (Google News) mentioning Google. Headlines link to the original; verify before acting.

Change history