FrontierMath-Tier-4-v2-Private
Integrity rank #2 of 65 · 65 models scored · top score 100 · GPT-6.1 Sol
Research-level mathematical reasoning on novel problems at the frontier of human mathematical expertise.
Strongest on discrimination, weakest on saturation headroom.
Reported scores — protocols may differ. How we compare
| # | Model | Score | Evidence | Measured |
|---|---|---|---|---|
| 1 | GPT-6.1 Sol | 100 | unverifiedT2 | 2026-09-29 |
| 2 | GPT-6 Astra | 97.6 | unverifiedT2 | 2026-09-03 |
| 3 | Claude Opus 5.5 | 95 | unverifiedT2 | 2026-09-22 |
| 4 | Claude Fable 5 | 90.2 | unverifiedT2 | 2026-06-09 |
| 5 | GPT-6 Sol | 90 | unverifiedT2 | 2026-09-22 |
| 6 | Claude Fable 5.1 | 87.8 | unverifiedT2 | 2026-09-01 |
| 7 | GPT-5.6 Sol | 82.93 | reproducedT1 | 2026-07-09 |
| 8 | Claude Sonnet 5.5 | 80.49 | unverifiedT2 | 2026-09-28 |
| 9 | Gemini 4 Argon | 78.05 | unverifiedT2 | 2026-09-30 |
| 10 | GPT-5.5 Pro | 78.05 | reproducedT1 | 2026-04-23 |
| 11 | Claude Opus 5 | 73.17 | reproducedT1 | 2026-07-24 |
| 12 | GPT-5.5 | 72.5 | reproducedT1 | 2026-04-23 |
| 13 | GPT-5.6 Terra | 70.73 | reproducedT1 | 2026-07-09 |
| 14 | GPT-5.6 Luna | 60.98 | reproducedT1 | 2026-07-09 |
| 15 | GPT-5.4 Pro | 58.54 | reproducedT1 | 2026-03-05 |
| 16 | GPT-6 Luna | 56.1 | unverifiedT2 | 2026-09-22 |
| 17 | Claude Opus 4.8 | 56.1 | reproducedT1 | 2026-05-28 |
| 18 | GPT-5.4 | 49 | reproducedT1 | 2026-03-05 |
| 19 | Claude Haiku 5.5 | 46.34 | unverifiedT2 | 2026-10-07 |
| 20 | Muse Spark 1.3 | 46.34 | unverifiedT2 | 2026-09-02 |
| 21 | Qwen 3.8 Max | 46.34 | reproducedT1 | 2026-08-02 |
| 22 | GPT-5.2 Pro | 46 | reproducedT1 | 2025-12-11 |
| 23 | Kimi K3 | 39.02 | reproducedT1 | 2026-07-16 |
| 24 | Gemini 3.7 Flash | 36.59 | reproducedT1 | 2026-08-13 |
| 25 | Qwen3.7-Max | 34.15 | reproducedT1 | 2026-05-19 |
| 26 | Qwen3.8 Max (0902) | 34.15 | unverifiedT2 | 2026-09-01 |
| 27 | Grok 4.6 | 31.71 | reproducedT1 | 2026-08-12 |
| 28 | Claude Opus 4.7 | 31.71 | reproducedT1 | 2026-04-16 |
| 29 | GPT-5.2 | 31.7 | reproducedT1 | 2025-12-11 |
| 30 | GLM-5.2 | 29.27 | reproducedT1 | 2026-06-16 |
| 31 | Claude Sonnet 5 | 29.27 | reproducedT1 | 2026-06-30 |
| 32 | GLM-5.3 | 29.27 | unverifiedT2 | 2026-08-14 |
| 33 | Gemini 3.1 Pro | 26.83 | reproducedT1 | 2026-02-19 |
| 34 | Gemini 3.5 Flash | 26.83 | reproducedT1 | 2026-05-19 |
| 35 | Claude Opus 4.6 | 26.83 | reproducedT1 | 2026-02-05 |
| 36 | DeepSeek V4 Pro 0813 | 26.83 | reproducedT1 | 2026-08-13 |
| 37 | DeepSeek V4.1 Flash | 26.83 | unverifiedT2 | 2026-09-09 |
| 38 | Kimi K2.6 | 25.64 | reproducedT1 | 2026-04-20 |
| 39 | Grok 4.5 | 24.39 | reproducedT1 | 2026-07-08 |
| 40 | DeepSeek V4 Flash 0731 | 24.39 | reproducedT1 | 2026-07-31 |
| 41 | Gemini 3.6 Flash | 21.95 | reproducedT1 | 2026-07-21 |
| 42 | Gemini 3.8 Flash | 21.95 | unverifiedT2 | 2026-09-02 |
| 43 | GPT-5 | 21.95 | reproducedT1 | 2025-08-07 |
| 44 | GPT-5 Pro | 19.51 | reproducedT1 | 2025-10-07 |
| 45 | Gemini 3 Flash | 17.07 | reproducedT1 | 2025-12-17 |
| 46 | Grok 4.20 | 17.07 | reproducedT1 | 2026-02-17 |
| 47 | Grok 4.7 | 17.07 | unverifiedT2 | 2026-09-21 |
| 48 | Inkling-Small | 17.07 | reproducedT1 | 2026-07-15 |
| 49 | GLM-5.3-Flash | 17.07 | unverifiedT2 | 2026-08-20 |
| 50 | Grok 4.3 Beta | 14.63 | reproducedT1 | 2026-04-17 |
| 51 | Kimi K2.7 Code | 12.2 | reproducedT1 | 2026-06-12 |
| 52 | GPT-5 mini | 12.2 | reproducedT1 | 2025-08-07 |
| 53 | GPT-5.4 Nano | 12.2 | reproducedT1 | 2026-03-17 |
| 54 | GPT-5.4 Mini | 9.76 | reproducedT1 | 2026-03-17 |
| 55 | Inkling | 4.88 | reproducedT1 | 2026-07-15 |
| 56 | Claude Opus 4.5 | 4.88 | reproducedT1 | 2025-11-24 |
| 57 | o4-mini | 4.88 | reproducedT1 | 2025-04-16 |
| 58 | DeepSeek-V4-Pro | 2.44 | reproducedT1 | 2026-04-24 |
| 59 | Claude Opus 4.1 | 2.44 | reproducedT1 | 2025-08-05 |
| 60 | Claude Sonnet 4.5 | 2.44 | reproducedT1 | 2025-09-29 |
| 61 | GPT-5 nano | 2.44 | reproducedT1 | 2025-08-07 |
| 62 | GPT-5.5 Instant | 2.44 | reproducedT1 | 2026-05-05 |
| 63 | Gemini 2.5 Pro (Jun 2025) | 0 | reproducedT1 | 2025-06-05 |
| 64 | Gemini 3.5 Flash-Lite | 0 | reproducedT1 | 2026-07-21 |
| 65 | o3-mini | 0 | reproducedT1 | 2025-01-31 |
Continue from this leaderboard into the tightest head-to-head reads.
Weighted composite. Each dimension opens to the source-backed sub-signals it is computed from.
Claims drawn from cited facts, not live model generation.
This benchmark tests research-level reasoning on novel mathematics problems.
1 cited fact
This benchmark has scores from 65 tracked models, with a recorded score range spanning 100.0.
2 cited facts
Its Benchmark Integrity Index score is 100, with grade A and rank 2 out of 65 tracked benchmarks.
4 cited facts
It is not flagged as saturated. The top recorded score is 100.0, with 0.0 points of headroom. The computed top-score cluster contains 1 models.
4 cited facts
The harness classification is consistent harness. Test-set privacy is recorded as held-out.
2 cited facts
FrontierMath-Tier-4-v2-Private: FrontierMath-Tier-4-v2-Private as reported in Epoch AI's Capabilities Index CSV.
GPT-6.1 Sol leads FrontierMath-Tier-4-v2-Private at 100. The full leaderboard above lists every recorded measurement, not just the headline number.
65 models have recorded scores on FrontierMath-Tier-4-v2-Private, spanning a score spread of 100.
tensor.news grades FrontierMath-Tier-4-v2-Private A for integrity (score 100/100), ranking #2 of 65 benchmarks we assess across discrimination, saturation, contamination resistance, harness comparability, and freshness.
No — FrontierMath-Tier-4-v2-Private still has headroom and continues to discriminate between models rather than bunching them at the ceiling.
Its test set is held-out, which lowers — but does not eliminate — the risk that training data leaked into the questions.
Scores are reported under a consistent harness, so comparisons on FrontierMath-Tier-4-v2-Private are reasonably apples-to-apples.
Every FrontierMath-Tier-4-v2-Private measurement is source-backed and tagged reproduced, self-reported, or unverified; the underlying record cites epoch.ai/benchmarks/frontiermath-tier-4-v2.