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

GPT-5.1 vs GPT-5.4

Verdict

GPT-5.4 leads 14–2 across 16 shared benchmarks.

GPT-5.1 2 · GPT-5.4 14 · higher isn't always better — see caveats.

Per-benchmark head-to-head

GPT-5.1GPT-5.4
  • 17.5
    APEX-Agents
    35.9
  • 72.8
    ARC-AGI
    93.67
  • 17.64
    ARC-AGI-2
    73.95
  • 32
    Chess Puzzles
    44
  • 23.7
    CL-bench
    27.9
  • 17.3
    CL-bench Life
    21.7
  • 4.86
    CritPt
    23.43
  • 42.79
    DeepResearch Bench
    35.09
  • 83.5
    GPQA diamond
    91.07
  • 13.73
    GSO-Bench
    31.37
  • 19.83
    HLE
    33.03
  • 88.6
    OTIS Mock AIME 2024-2025
    95.3
  • 48.9
    SimpleQA Verified
    44.83
  • 67.98
    SWE-Bench verified
    76.86
  • 47.6
    Terminal Bench
    81.8
  • 60.77
    WeirdML
    77.7
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

  • GPT-5.188.6
  • GPT-5.495.3