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

GeoBench

GeoBench

Rank #19 of 51 · 26 models scored · top score 88

AINTEGRITY 89 / 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

12 headroomtop 88ceiling 100

Score distribution

median 71Gemini 3 Flash 88316191

State of the art over time

3161918820242025

Integrity breakdown

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

Discrimination30%100

Does it still separate models?

Saturation headroom30%93

How far from ceiling / clustered at the top?

Contamination resistance15%42

Public vs held-out; training-leak risk.

Harness comparability15%100

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

Freshness10%93

How old is the benchmark?

Leaderboard

#ModelScoreEvidenceMeasured
1Gemini 3 Flash88unverified2025-12-17
2Gemini 2.5 Pro (May 2025)86unverified2025-05-06
3Gemini 3 Pro84unverified2025-11-18
4GPT-581unverified2025-08-07
5Gemini 2.5 Pro (Mar 2025)81unverified2025-03-25
6o180unverified2024-12-05
7Gemini 2.0 Flash (Feb 2025)77unverified2024-12-11
8Gemini 1.5 Flash (Sep 2024)76unverified2024-05-10
9Gemini 2.5 Flash (May 2025)76unverified2025-05-20
10Claude Opus 4.575unverified2025-11-24
11o374unverified2024-12-20
12Gemini 2.5 Flash (Apr 2025)73unverified2025-04-17
13GPT-4.172unverified2025-04-14
14GPT-4o (Nov 2024)71unverified2024-05-13
15Claude 3.7 Sonnet68unverified2025-02-24
16o4-mini64unverified2025-04-16
17GPT-4o mini64unverified2024-07-18
18Claude 3.5 Sonnet (October 2024)62unverified2024-10-22
19Qwen2.5-72B62unverified2024-09-19
20Gemma 3 27B52unverified2025-03-12
21Llama 3.2 90B52unverified2024-09-24
22Llama 4 Maverick52unverified2025-04-05
23Claude Opus 449unverified2025-05-22
24Grok 445unverified2025-07-09
25Claude Sonnet 437unverified2025-05-22
26Claude 3.5 Haiku34unverified2024-10-22

At a glance

Models scored
26
Top model
Gemini 3 Flash
Top score
88
Score spread
54
Headroom
12

Benchmark context

Task type
multimodal
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
2025-03-01
License
CC BY 4.0
Source
epoch.ai/data/eci_benchmarks.csv

Provenance

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

Follow the record

Integrity neighbors

  • GSO-BenchA · 90
  • BalrogA · 89
  • ExploitBenchA · 89
  • PostTrainBenchA · 89
  • OTIS Mock AIME 2024-2025A · 88
  • WeirdMLA · 88

Top models here

  • Gemini 3 Flash88
  • Gemini 2.5 Pro (May 2025)86
  • Gemini 3 Pro84
  • GPT-581
  • Gemini 2.5 Pro (Mar 2025)81

Head-to-head

  • Gemini 3 Flash vs Gemini 2.5 Pro (May 2025)
  • Gemini 3 Flash vs Gemini 3 Pro
  • Gemini 2.5 Pro (May 2025) vs Gemini 3 Pro