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

Lech Mazur Writing

Lech Mazur Writing

Rank #37 of 51 · 39 models scored · top score 87.29

BINTEGRITY 76 / 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

12.7 headroomtop 87.3ceiling 100

Score distribution

median 77.3Kimi K2 (Sep 2025) 87.3597489

State of the art over time

59748987.320242025

Integrity breakdown

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

Discrimination30%67

Does it still separate models?

Saturation headroom30%84

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%91

How old is the benchmark?

Leaderboard

#ModelScoreEvidenceMeasured
1Kimi K2 (Sep 2025)87.29unverified2025-07-11
2GPT-587.23unverified2025-08-07
3Qwen3-Max87.11unverified2025-09-05
4Kimi K2 (Jul 2025)86.93unverified2025-07-11
5o3-pro86.28unverified2025-06-10
6Gemini 2.5 Pro (Jun 2025)86.02unverified2025-06-05
7Claude Opus 4.185.38unverified2025-08-05
8DeepSeek-V3.185.17unverified2025-08-21
9Qwen3-235B-A22B-Thinking (Jul 2025)84.95unverified2025-07-25
10o383.9unverified2024-12-20
11Qwen3-235B-A22B83unverified2025-04-28
12DeepSeek-R183unverified2025-01-20
13GPT-4o (Nov 2024)81.8unverified2024-05-13
14Grok 4 Fast81.14unverified2025-09-19
15Claude 3.7 Sonnet81.1unverified2025-02-24
16Grok 480.68unverified2025-07-09
17Gemini 2.5 Pro (Mar 2025)80.5unverified2025-03-25
18Claude 3.5 Sonnet (October 2024)80.3unverified2024-10-22
19Gemma 3 27B79.9unverified2025-03-12
20gpt-oss-120b77.26unverified2025-08-05
21DeepSeek-V3 (Mar 2025)77unverified2025-03-24
22Gemini 2.5 Flash (Apr 2025)76.5unverified2025-04-17
23Grok 376.4unverified2025-02-17
24GPT-4.575.6unverified2025-02-27
25o4-mini75unverified2025-04-16
26Gemini 2.0 Flash Thinking (Jan 2025)73.8unverified2025-01-21
27Grok-3 mini73.5unverified2025-02-19
28Claude 3.5 Haiku73.5unverified2024-10-22
29Qwen2.5-Max72.9unverified2025-01-25
30Gemini 2.0 Flash (Dec 2024)71.5unverified2024-12-11
31o170.2unverified2024-12-05
32Mistral Large 2 (Jul 2024)69unverified2024-07-24
33GPT-4o mini67.2unverified2024-07-18
34o1-mini64.9unverified2024-09-12
35Llama 4 Maverick63.7unverified2025-04-05
36Grok-2 (Dec 2024)63.6unverified2024-08-13
37Phi-462.6unverified2024-12-12
38o3-mini61.7unverified2025-01-31
39Amazon Nova Pro60.5unverified2024-12-03

At a glance

Models scored
39
Top model
Kimi K2 (Sep 2025)
Top score
87.29
Score spread
26.79
Headroom
12.71

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
2025-01-31
License
CC BY 4.0
Source
epoch.ai/data/eci_benchmarks.csv

Provenance

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

Follow the record

Integrity neighbors

  • OSWorldB · 80
  • CybenchB · 79
  • DeepResearch BenchB · 78
  • ARC-AGIB · 74
  • MATH level 5B · 73
  • BBHB · 71

Top models here

  • Kimi K2 (Sep 2025)87.29
  • GPT-587.23
  • Qwen3-Max87.11
  • Kimi K2 (Jul 2025)86.93
  • o3-pro86.28

Head-to-head

  • Kimi K2 (Sep 2025) vs GPT-5
  • Kimi K2 (Sep 2025) vs Qwen3-Max
  • GPT-5 vs Qwen3-Max