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 2.0

OSWorld 2.0

Rank #32 of 51 · 5 models scored · top score 20.6

BINTEGRITY 81 / 100

unknown

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

consistent harnesspublic test set

Headroom to ceiling

79.4 headroomtop 20.6ceiling 100

Score distribution

median 13Claude Opus 4.8 20.6312.522

State of the art over time

3132220.620262026

Integrity breakdown

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

Discrimination30%40

Does it still separate models?

Saturation headroom30%96

How far from ceiling / clustered at the top?

Contamination resistance15%100

Public vs held-out; training-leak risk.

Harness comparability15%100

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

Freshness10%100

How old is the benchmark?

Leaderboard

#ModelScoreEvidenceMeasured
1Claude Opus 4.820.6unverified2026-05-28
2Claude Opus 4.718.2unverified2026-04-16
3GPT-5.513unverified2026-04-23
4Claude Sonnet 4.69.3unverified2026-02-17
5Kimi K2.64.6unverified2026-04-20

At a glance

Models scored
5
Top model
Claude Opus 4.8
Top score
20.6
Score spread
16
Headroom
79.4

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
unknown
License
CC BY 4.0
Source
epoch.ai/data/eci_benchmarks.csv

Provenance

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

Follow the record

Integrity neighbors

  • CL-benchB · 82
  • CL-bench LifeB · 81
  • GPQA diamondB · 81
  • Remote Labor IndexB · 81
  • OSWorldB · 80
  • CybenchB · 79

Top models here

  • Claude Opus 4.820.6
  • Claude Opus 4.718.2
  • GPT-5.513
  • Claude Sonnet 4.69.3
  • Kimi K2.64.6

Head-to-head

  • Claude Opus 4.8 vs Claude Opus 4.7
  • Claude Opus 4.8 vs GPT-5.5
  • Claude Opus 4.7 vs GPT-5.5