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.

← Integrity Index

GSO-Bench

GSO-Bench

Rank #16 of 51 · 22 models scored · top score 44.12

AINTEGRITY 90 / 100

unknown

Strongest on discrimination, weakest on contamination resistance. Caveat: the public test set leaves contamination on the table.

consistent harnesspublic test set

Headroom to ceiling

55.9 headroomtop 44.1ceiling 100

Score distribution

median 8.8Claude Opus 4.7 44.102346

State of the art over time

0234644.120242026

Integrity breakdown

Weighted composite. Each signal is computed from the source-backed leaderboard below.

Discrimination30%100

Does it still separate models?

Saturation headroom30%98

How far from ceiling / clustered at the top?

Contamination resistance15%36

Public vs held-out; training-leak risk.

Harness comparability15%100

Apples-to-apples, or mixed / vendor-optimized?

Freshness10%98

How old is the benchmark?

Leaderboard

#ModelScoreEvidenceMeasured
1Claude Opus 4.744.12unverified2026-04-16
2Claude Opus 4.641.2unverified2026-02-05
3GPT-5.540.2unverified2026-04-23
4GPT-5.431.37unverified2026-03-05
5GPT-5.227.4unverified2025-12-11
6Claude Opus 4.526.5unverified2025-11-24
7Gemini 3.1 Pro22.55unverified2026-02-19
8Gemini 3 Pro18.6unverified2025-11-18
9Claude Sonnet 4.514.71unverified2025-09-29
10GPT-5.113.73unverified2025-11-13
11Gemini 3 Flash9.8unverified2025-12-17
12o38.8unverified2024-12-20
13Claude Opus 46.9unverified2025-05-22
14GPT-56.9unverified2025-08-07
15Kimi K2 (Jul 2025)4.9unverified2025-07-11
16Claude Sonnet 44.9unverified2025-05-22
17Claude 3.5 Sonnet (October 2024)4.6unverified2024-10-22
18Gemini 2.5 Pro (Jun 2025)3.92unverified2025-06-05
19Claude 3.7 Sonnet3.8unverified2025-02-24
20o4-mini3.6unverified2025-04-16
21o3-mini1.3unverified2025-01-31
22GPT-4o (Nov 2024)0unverified2024-05-13

At a glance

Models scored
22
Top model
Claude Opus 4.7
Top score
44.12
Score spread
44.12
Headroom
55.88

Benchmark context

Task type
unknown
Metric
normalized performance in Epoch CSV, rendered as accuracy (%)
Score ceiling
100
Human baseline
unknown
Test-set privacy
unknown
Construction
unknown
Key assumptions
unknown
Limitations
unknown
Contamination history
unknown
Harness notes
Epoch CSV reports model, model version, performance, optimized flag, score source, benchmark release date, and model date.
Authors
unknown
Released
2025-05-29
License
CC BY 4.0
Source
epoch.ai/data/eci_benchmarks.csv

Provenance

22 source records · scores from epoch_benchmarks. Every fact on this page is evidence-linked.

Follow the record

Integrity neighbors

  • FrontierMath-Tier-4-2025-07-01-PrivateA · 93
  • CritPtA · 92
  • The Agent CompanyA · 92
  • BalrogA · 89
  • ExploitBenchA · 89
  • GeoBenchA · 89

Top models here

  • Claude Opus 4.744.12
  • Claude Opus 4.641.2
  • GPT-5.540.2
  • GPT-5.431.37
  • GPT-5.227.4

Head-to-head

  • Claude Opus 4.7 vs Claude Opus 4.6
  • Claude Opus 4.7 vs GPT-5.5
  • Claude Opus 4.6 vs GPT-5.5