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.

← Compare

Gemini 2.5 Pro (Jun 2025) vs GPT-5

Verdict

GPT-5 leads 21–2 across 23 shared benchmarks.

Gemini 2.5 Pro (Jun 2025) 2 · GPT-5 21 · higher isn't always better — see caveats.

Per-benchmark head-to-head

Gemini 2.5 Pro (Jun 2025)GPT-5
  • 83.1
    Aider polyglot
    88
  • 6.6
    APEX-Agents
    18.3
  • 41
    ARC-AGI
    65.7
  • 4.86
    ARC-AGI-2
    9.86
  • 20
    Chess Puzzles
    37
  • 2
    CritPt
    12.6
  • 49.71
    DeepResearch Bench
    55.13
  • 91.7
    Fiction.LiveBench
    97.2
  • 0
    FrontierMath-Tier-4-v2-Private
    21.95
  • 24.56
    FrontierMath-Tiers-1-3-v2-Private
    55.44
  • 23.3
    GDPval
    34.8
  • 80.39
    GPQA diamond
    81.57
  • 3.92
    GSO-Bench
    6.9
  • 17.69
    HLE
    21.55
  • 86.02
    Lech Mazur Writing
    87.23
  • 84.71
    OTIS Mock AIME 2024-2025
    91.38
  • 0.83
    Remote Labor Index
    1.67
  • 54.88
    SimpleBench
    48.04
  • 56
    SimpleQA Verified
    50.6
  • 57.56
    SWE-Bench verified
    73.55
  • 32.6
    Terminal Bench
    49.6
  • 19.6
    VPCT
    49
  • 54.03
    WeirdML
    60.7
A higher number is not always a better model. Each score is task performance under a disclosed harness, and some of these benchmarks are saturated or scored under mixed harnesses. Follow any benchmark to its integrity grade before reading a win as decisive.

Follow the record

These models

  • Gemini 2.5 Pro (Jun 2025)91.7
  • GPT-598.13