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

OSWorld

OSWorld

Rank #34 of 51 · 8 models scored · top score 72.1

BINTEGRITY 80 / 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

27.9 headroomtop 72.1ceiling 100

Score distribution

median 43.9Claude Sonnet 4.6 72.1238.575

State of the art over time

2397572.120242026

Integrity breakdown

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

Discrimination30%100

Does it still separate models?

Saturation headroom30%91

How far from ceiling / clustered at the top?

Contamination resistance15%0

Public vs held-out; training-leak risk.

Harness comparability15%100

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

Freshness10%75

How old is the benchmark?

Leaderboard

#ModelScoreEvidenceMeasured
1Claude Sonnet 4.672.1unverified· optimized2026-02-17
2Claude Opus 4.566.3unverified· optimized2025-11-24
3Kimi K2.563.3unverified· optimized2026-02-02
4Claude Sonnet 4.562.9unverified· optimized2025-09-29
5Claude Sonnet 443.9unverified· optimized2025-05-22
6Claude 3.7 Sonnet35.8unverified· optimized2025-02-24
7o323unverified· optimized2024-12-20
8Qwen2.5-72B5unverified· optimized2024-09-19

At a glance

Models scored
8
Top model
Claude Sonnet 4.6
Top score
72.1
Score spread
67.1
Headroom
27.9

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
2024-04-11
License
CC BY 4.0
Source
epoch.ai/data/eci_benchmarks.csv

Provenance

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

Follow the record

Integrity neighbors

  • GPQA diamondB · 81
  • OSWorld 2.0B · 81
  • Remote Labor IndexB · 81
  • CybenchB · 79
  • DeepResearch BenchB · 78
  • Lech Mazur WritingB · 76

Top models here

  • Claude Sonnet 4.672.1
  • Claude Opus 4.566.3
  • Kimi K2.563.3
  • Claude Sonnet 4.562.9
  • Claude Sonnet 443.9

Head-to-head

  • Claude Sonnet 4.6 vs Claude Opus 4.5
  • Claude Sonnet 4.6 vs Kimi K2.5
  • Claude Opus 4.5 vs Kimi K2.5