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

The Agent Company

The Agent Company

Rank #15 of 51 · 14 models scored · top score 42.9

AINTEGRITY 92 / 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

57.1 headroomtop 42.9ceiling 100

Score distribution

median 8.6DeepSeek-V3.2-Exp 42.9022.545

State of the art over time

0234542.920242025

Integrity breakdown

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

Discrimination30%100

Does it still separate models?

Saturation headroom30%94

How far from ceiling / clustered at the top?

Contamination resistance15%64

Public vs held-out; training-leak risk.

Harness comparability15%100

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

Freshness10%89

How old is the benchmark?

Leaderboard

#ModelScoreEvidenceMeasured
1DeepSeek-V3.2-Exp42.9unverified2025-09-29
2Gemini 2.5 Flash (Sep 2025)41.1unverified2025-09-25
3Claude Sonnet 433.1unverified2025-05-22
4Claude 3.7 Sonnet30.9unverified2025-02-24
5Gemini 2.5 Pro (May 2025)30.3unverified2025-05-06
6Claude 3.5 Sonnet (October 2024)24unverified2024-10-22
7Gemini 2.0 Flash (Feb 2025)11.4unverified2024-12-11
8GPT-4o (Nov 2024)8.6unverified2024-05-13
9Llama 3.1-405B7.4unverified2024-07-23
10Llama 3.1-70B6.9unverified2024-07-23
11Qwen2.5-72B5.7unverified2024-09-19
12Gemini 1.5 Pro (Sept 2024)3.4unverified2024-09-24
13Amazon Nova Pro1.7unverified2024-12-03
14Qwen2-72B1.1unverified2024-06-07

At a glance

Models scored
14
Top model
DeepSeek-V3.2-Exp
Top score
42.9
Score spread
41.8
Headroom
57.1

Benchmark context

Task type
agentic
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
2024-12-18
License
CC BY 4.0
Source
epoch.ai/data/eci_benchmarks.csv

Provenance

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

Follow the record

Integrity neighbors

  • GDPvalA · 98
  • FrontierMath-Tier-4-2025-07-01-PrivateA · 93
  • CritPtA · 92
  • GSO-BenchA · 90
  • BalrogA · 89
  • ExploitBenchA · 89

Top models here

  • DeepSeek-V3.2-Exp42.9
  • Gemini 2.5 Flash (Sep 2025)41.1
  • Claude Sonnet 433.1
  • Claude 3.7 Sonnet30.9
  • Gemini 2.5 Pro (May 2025)30.3

Head-to-head

  • DeepSeek-V3.2-Exp vs Gemini 2.5 Flash (Sep 2025)
  • DeepSeek-V3.2-Exp vs Claude Sonnet 4
  • Gemini 2.5 Flash (Sep 2025) vs Claude Sonnet 4