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

FrontierMath-Tier-4-v2-Private

FrontierMath-Tier-4-v2-Private

Rank #7 of 51 · 31 models scored · top score 87.8

AINTEGRITY 99 / 100

unknown

Strongest on discrimination, weakest on saturation headroom.

consistent harnessheld-out test set

Headroom to ceiling

12.2 headroomtop 87.8ceiling 100

Score distribution

median 21.9Claude Fable 5 87.804692

State of the art over time

0469287.820252026

Integrity breakdown

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

Discrimination30%100

Does it still separate models?

Saturation headroom30%97

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 Fable 587.8reproduced2026-06-09
2GPT-5.5 Pro78.05reproduced2026-04-23
3GPT-5.572.5reproduced2026-04-23
4GPT-5.4 Pro58.54reproduced2026-03-05
5Claude Opus 4.856.1reproduced2026-05-28
6GPT-5.449reproduced2026-03-05
7GPT-5.2 Pro46reproduced2025-12-11
8Qwen3.7-Max34.15reproduced2026-05-19
9Claude Opus 4.731.71reproduced2026-04-16
10GPT-5.231.7reproduced2025-12-11
11GLM-5.229.27reproduced2026-06-16
12Gemini 3.1 Pro26.83reproduced2026-02-19
13Gemini 3.5 Flash26.83reproduced2026-05-19
14Claude Opus 4.626.83reproduced2026-02-05
15Kimi K2.625.64reproduced2026-04-20
16GPT-521.95reproduced2025-08-07
17GPT-5 Pro19.51reproduced2025-10-07
18Gemini 3 Flash17.07reproduced2025-12-17
19Grok 4.3 Beta14.63reproduced2026-04-17
20Kimi K2.7 Code12.2reproduced2026-06-12
21GPT-5 mini12.2reproduced2025-08-07
22GPT-5.4 Nano12.2reproduced2026-03-17
23GPT-5.4 Mini9.76reproduced2026-03-17
24Claude Opus 4.54.88reproduced2025-11-24
25o4-mini4.88reproduced2025-04-16
26Claude Opus 4.12.44reproduced2025-08-05
27Claude Sonnet 4.52.44reproduced2025-09-29
28DeepSeek-V4-Pro2.44reproduced2026-04-24
29GPT-5 nano2.44reproduced2025-08-07
30o3-mini0reproduced2025-01-31
31Gemini 2.5 Pro (Jun 2025)0reproduced2025-06-05

At a glance

Models scored
31
Top model
Claude Fable 5
Top score
87.8
Score spread
87.8
Headroom
12.2

Benchmark context

Task type
math
Metric
normalized performance in Epoch CSV, rendered as accuracy (%)
Score ceiling
100
Human baseline
unknown
Test-set privacy
held-out
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

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

Follow the record

Integrity neighbors

  • SimpleQA VerifiedA · 100
  • ARC-AGI-2A · 99
  • FrontierMath-2025-02-28-PrivateA · 99
  • CadEvalA · 98
  • CursorBenchA · 98
  • FrontierMath-Tiers-1-3-v2-PrivateA · 98

Top models here

  • Claude Fable 587.8
  • GPT-5.5 Pro78.05
  • GPT-5.572.5
  • GPT-5.4 Pro58.54
  • Claude Opus 4.856.1

Head-to-head

  • Claude Fable 5 vs GPT-5.5 Pro
  • Claude Fable 5 vs GPT-5.5
  • GPT-5.5 Pro vs GPT-5.5