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

Balrog

Balrog

Rank #17 of 51 · 24 models scored · top score 58.1

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

41.9 headroomtop 58.1ceiling 100

Score distribution

median 29.5Gemini 3 Pro 58.1934.560

State of the art over time

9356058.120242026

Integrity breakdown

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

Discrimination30%100

Does it still separate models?

Saturation headroom30%95

How far from ceiling / clustered at the top?

Contamination resistance15%46

Public vs held-out; training-leak risk.

Harness comparability15%100

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

Freshness10%88

How old is the benchmark?

Leaderboard

#ModelScoreEvidenceMeasured
1Gemini 3 Pro58.1unverified2025-11-18
2Gemini 3.1 Pro57unverified2026-02-19
3Gemini 3 Flash48.1unverified2025-12-17
4Grok 443.6unverified2025-07-09
5Claude Opus 4.543.5unverified2025-11-24
6Gemini 2.5 Pro (Mar 2025)43.3unverified2025-03-25
7DeepSeek-R134.9unverified2025-01-20
8Gemini 2.5 Flash (Jun 2025)33.5unverified2025-06-17
9GPT-532.8unverified2025-08-07
10Claude 3.5 Sonnet (October 2024)32.6unverified2024-10-22
11GPT-4o (May 2024)32.3unverified2024-05-13
12Claude Haiku 4.531.2unverified2025-10-15
13Grok 329.5unverified2025-02-17
14Llama 3.1-70B27.9unverified2024-07-23
15Llama 3.2 90B27.3unverified2024-09-24
16Llama 3.3 70B23unverified2024-12-06
17Gemini 1.5 Pro (Sept 2024)21unverified2024-09-24
18Claude 3.5 Haiku19.3unverified2024-10-22
19Mistral NeMo17.6unverified2024-07-18
20GPT-4o mini17.4unverified2024-07-18
21Qwen2.5-72B16.2unverified2024-09-19
22Llama 3.1-8B15.1unverified2024-07-23
23Gemini 1.5 Flash (Sep 2024)14.6unverified2024-05-10
24Phi-411.6unverified2024-12-12

At a glance

Models scored
24
Top model
Gemini 3 Pro
Top score
58.1
Score spread
46.5
Headroom
41.9

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
2024-11-20
License
CC BY 4.0
Source
epoch.ai/data/eci_benchmarks.csv

Provenance

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

Follow the record

Integrity neighbors

  • CritPtA · 92
  • The Agent CompanyA · 92
  • GSO-BenchA · 90
  • ExploitBenchA · 89
  • GeoBenchA · 89
  • PostTrainBenchA · 89

Top models here

  • Gemini 3 Pro58.1
  • Gemini 3.1 Pro57
  • Gemini 3 Flash48.1
  • Grok 443.6
  • Claude Opus 4.543.5

Head-to-head

  • Gemini 3 Pro vs Gemini 3.1 Pro
  • Gemini 3 Pro vs Gemini 3 Flash
  • Gemini 3.1 Pro vs Gemini 3 Flash