tensor.news
Integrity IndexModelsCompareLabsPapersMethodologyAbout
tensor.news

© 2026

Evidence

  • Integrity Index
  • Model Claim Ledger
  • Compare

Entities

  • Models
  • Benchmarks
  • Labs
  • Papers

Method

  • Methodology
  • About
  • Privacy

Machine

  • Feed (Atom)
  • llms.txt
  • llms-full.txt
  • Sitemap

tensor.news grades benchmarks and tags model claims by evidence status. Every fact carries a source record and provenance. Scores are framed as task performance under a disclosed harness, never as deployed capability.

← Models

Gemini 3.1 Pro

Google DeepMind·released 2026-02-19

independently reproduced

Scored on 22 benchmarks, leading 3. 32% of its scores are independently reproduced — the rest are self-reported or unverified.

22
Benchmarks scored
3
SOTA leads
98
Best score
32%
Reproduced
7 of 22

Leads on: ARC-AGI, HLE, SimpleQA Verified

Benchmark fingerprint

Every score, ranked, colored by evidence status.

  • ARC-AGI
    98
  • OTIS Mock AIME 2024-2025
    95.6
  • GPQA diamond
    92.13
  • Terminal Bench
    80.2
  • SimpleQA Verified
    77.3
  • ARC-AGI-2
    77.1
  • SWE-Bench verified
    75.62
  • SimpleBench
    75.52
  • WeirdML
    72.1
  • FrontierMath-Tiers-1-3-v2-Private
    59.65
  • Balrog
    57
  • Chess Puzzles
    55
  • HLE
    43.74
  • APEX-Agents
    33.5
  • FrontierMath-Tier-4-v2-Private
    26.83
  • ExploitBench
    26
  • GSO-Bench
    22.55
  • PostTrainBench
    21.59
  • CL-bench
    20.8
  • CritPt
    17.71
  • CL-bench Life
    16.9
  • GBAEval
    0.85

All benchmark scores

BenchmarkScoreEvidence statusMeasured
ARC-AGI98accuracy (%)unverified· optimized2026-02-19
OTIS Mock AIME 2024-202595.6accuracy (%)reproduced· optimized2026-02-19
GPQA diamond92.13accuracy (%)reproduced· optimized2026-02-19
Terminal Bench80.2accuracy (%)unverified· optimized2026-02-19
SimpleQA Verified77.3accuracy (%)reproduced2026-02-19
ARC-AGI-277.1accuracy (%)unverified2026-02-19
SWE-Bench verified75.62accuracy (%)reproduced· optimized2026-02-19
SimpleBench75.52accuracy (%)unverified2026-02-19
WeirdML72.1accuracy (%)unverified2026-02-19
FrontierMath-Tiers-1-3-v2-Private59.65accuracy (%)reproduced2026-02-19
Balrog57accuracy (%)unverified2026-02-19
Chess Puzzles55accuracy (%)reproduced2026-02-19
HLE43.74accuracy (%)unverified2026-02-19
APEX-Agents33.5accuracy (%)unverified2026-02-19
FrontierMath-Tier-4-v2-Private26.83accuracy (%)reproduced2026-02-19
ExploitBench26accuracy (%)unverified2026-02-19
GSO-Bench22.55accuracy (%)unverified2026-02-19
PostTrainBench21.59accuracy (%)unverified2026-02-19
CL-bench20.8accuracy (%)unverified2026-02-19
CritPt17.71accuracy (%)unverified2026-02-19
CL-bench Life16.9accuracy (%)unverified2026-02-19
GBAEval0.85accuracy (%)unverified2026-02-19
22 source records · epoch_benchmarks. Every score is evidence-linked.

Follow the record

More from Google DeepMind

  • Gemini 2.5 Pro (May 2025)95.9
  • Gemini 2.5 Pro (Mar 2025)95.56
  • Gemini 3.5 Flash95.55
  • Gemini 3 Flash92.77
  • Gemini 2.5 Pro (Jun 2025)91.7
  • Gemini 3 Pro91.38

Head-to-head

  • vs PowerMoE-3b
  • vs Claude Fable 5
  • vs GPT-5.5
  • vs Llama 3.1-405B
  • vs phi-3-small 7.4B
  • vs DeepSeek-V3

Benchmarks it leads

  • ARC-AGI
  • HLE
  • SimpleQA Verified