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

DeepResearch Bench

DeepResearch Bench

Rank #36 of 51 · 22 models scored · top score 55.31

BINTEGRITY 78 / 100

unknown

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

consistent harnesspublic test set

Headroom to ceiling

44.7 headroomtop 55.3ceiling 100

Score distribution

median 46.6Claude Opus 4.6 55.32842.557

State of the art over time

28435755.320242026

Integrity breakdown

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

Discrimination30%65

Does it still separate models?

Saturation headroom30%90

How far from ceiling / clustered at the top?

Contamination resistance15%41

Public vs held-out; training-leak risk.

Harness comparability15%100

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

Freshness10%99

How old is the benchmark?

Leaderboard

#ModelScoreEvidenceMeasured
1Claude Opus 4.655.31unverified2026-02-05
2GPT-555.13unverified2025-08-07
3Claude Sonnet 4.654.87unverified2026-02-17
4GPT-5.554.01unverified2026-04-23
5Claude Sonnet 4.552.6unverified2025-09-29
6Claude Opus 4.850.23unverified2026-05-28
7Gemini 2.5 Pro (Jun 2025)49.71unverified2025-06-05
8Claude Opus 4.149.7unverified2025-08-05
9Claude Opus 449unverified2025-05-22
10Grok 447.9unverified2025-07-09
11Claude Sonnet 447.8unverified2025-05-22
12o346.6unverified2024-12-20
13Claude 3.7 Sonnet43.6unverified2025-02-24
14GPT-5.142.79unverified2025-11-13
15GPT-5.241.12unverified2025-12-11
16Gemini 3.1 Flash-Lite36.39unverified2026-03-03
17GPT-5.4 Mini36.26unverified2026-03-17
18DeepSeek-R1 (May 2025)35.1unverified2025-05-28
19GPT-5.435.09unverified2026-03-05
20Gemini 2.5 Pro (May 2025)31.9unverified2025-05-06
21GPT-4.129.31unverified2025-04-14
22Gemini 2.5 Flash (Apr 2025)29.19unverified2025-04-17

At a glance

Models scored
22
Top model
Claude Opus 4.6
Top score
55.31
Score spread
26.12
Headroom
44.69

Benchmark context

Task type
reasoning
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
2025-06-13
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

  • Remote Labor IndexB · 81
  • OSWorldB · 80
  • CybenchB · 79
  • Lech Mazur WritingB · 76
  • ARC-AGIB · 74
  • MATH level 5B · 73

Top models here

  • Claude Opus 4.655.31
  • GPT-555.13
  • Claude Sonnet 4.654.87
  • GPT-5.554.01
  • Claude Sonnet 4.552.6

Head-to-head

  • Claude Opus 4.6 vs GPT-5
  • Claude Opus 4.6 vs Claude Sonnet 4.6
  • GPT-5 vs Claude Sonnet 4.6