Benchmark
FrontierMath Tier 4 (v2)
Math benchmarks measure how well models solve maths problems. Scores are percentages. Higher is better.
- Models measured
- 74
- Scores recorded
- 104
- Sources
- 3
- IQ weight
- 2.74
Score distribution
- 14.5%
- Mean
- 0.561%
- Median
- 29.1%
- Std dev
- 0.049%
- 10th percentile
- 72.5%
- 90th percentile
- 0%
- Min
This benchmark has a weight of 2.74 for chat models, measured over 74 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model | Score |
|---|---|---|
| 1 | Claude Fable 5.2max· Epoch AI Benchmarking Hub data | 100% |
| 2 | GPT-6.1 Sol· LLM Stats data | 100% |
| 3 | GPT-6.1 Solmax· Epoch AI Benchmarking Hub data | 100% |
| 4 | GPT-6.2 Astramax· Epoch AI Benchmarking Hub data | 100% |
| 5 | GPT-6 Astra· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 97.6% |
| 6 | GPT-6 Astrahigh· Epoch AI Benchmarking Hub data | 97.6% |
| 7 | GPT-6 Astramax· Epoch AI Benchmarking Hub data | 97.6% |
| 8 | GPT-6 Astrav2 · GPT-6 Astra launch snapshot · 2026-09-01· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 97.6% |
| 9 | GPT-6 Astraxhigh· Epoch AI Benchmarking Hub data | 97.6% |
| 10 | Gemini 4 Neonxhigh· Epoch AI Benchmarking Hub data | 97.6% |
| 11 | GPT-6 Astramedium· Epoch AI Benchmarking Hub data | 97.6% |
| 12 | Gemini 4 Carbonhigh· Epoch AI Benchmarking Hub data | 95.1% |
| 13 | Claude Opus 5.5max· Epoch AI Benchmarking Hub data | 95% |
| 14 | GPT-6.1 Terramax· Epoch AI Benchmarking Hub data | 92.7% |
| 15 | Claude Fable 5max· Epoch AI Benchmarking Hub data | 90.2% |
| 16 | Claude Fable 5v2 · GPT-6 Astra launch snapshot · 2026-09-01· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 90.2% |
| 17 | GPT-6 Solmax· Epoch AI Benchmarking Hub data | 90% |
| 18 | Claude Fable 5.1max· Epoch AI Benchmarking Hub data | 87.8% |
| 19 | Gemini 4 Hydrogenxhigh· Epoch AI Benchmarking Hub data | 87.8% |
| 20 | GPT-6 Astralow· Epoch AI Benchmarking Hub data | 87.8% |
| 21 | Claude Fable 5v2 · GPT-5.6 launch snapshot · 2026-07-09· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 87.8% |
| 22 | Claude Fable 5.1v2 · GPT-6 Astra launch snapshot · 2026-09-01· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 87.8% |
| 23 | GPT-5.6 Sol· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 83% |
| 24 | GPT-5.6 Solv2 · GPT-5.6 launch snapshot · 2026-07-09· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 83% |
| 25 | GPT-5.6 Solv2 · GPT-6 Astra launch snapshot · 2026-09-01· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 83% |
| 26 | GPT-5.6 Solmax· Epoch AI Benchmarking Hub data | 82.9% |
| 27 | GPT-6 Astranone· Epoch AI Benchmarking Hub data | 82.9% |
| 28 | Claude Sonnet 5.5max· Epoch AI Benchmarking Hub data | 80.5% |
| 29 | GPT-5.6 Solpro / max· Epoch AI Benchmarking Hub data | 80.5% |
| 30 | Gemini 4 Argonhigh· Epoch AI Benchmarking Hub data | 78% |
| 31 | GPT-5.5 Proxhigh· Epoch AI Benchmarking Hub data | 78% |
| 32 | Claude Opus 5v2 · GPT-6 Astra launch snapshot · 2026-09-01· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 73.2% |
| 33 | Claude Opus 5max· Epoch AI Benchmarking Hub data | 73.2% |
| 34 | GPT 5.5v2 · GPT-5.6 launch snapshot · 2026-07-09· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 72.5% |
| 35 | GPT 5.5xhigh· Epoch AI Benchmarking Hub data | 72.5% |
| 36 | GPT-5.6 Terramax· Epoch AI Benchmarking Hub data | 70.7% |
| 37 | Muse Wonder 1.0max· Epoch AI Benchmarking Hub data | 70.7% |
| 38 | GPT-5.6 Terra· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 68.3% |
| 39 | GPT-5.6 Terrav2 · GPT-5.6 launch snapshot · 2026-07-09· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 68.3% |
| 40 | GPT-5.6 Lunamax· Epoch AI Benchmarking Hub data | 61% |
| 41 | GPT-5.4 Proxhigh· Epoch AI Benchmarking Hub data | 58.5% |
| 42 | GPT-5.6 Luna· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 58.5% |
| 43 | GPT-5.6 Lunav2 · GPT-5.6 launch snapshot · 2026-07-09· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 58.5% |
| 44 | Claude Opus 4.8v2 · GPT-5.6 launch snapshot · 2026-07-09· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 56.1% |
| 45 | Claude Opus 4.8max· Epoch AI Benchmarking Hub data | 56.1% |
| 46 | GPT-6 Lunamax· Epoch AI Benchmarking Hub data | 56.1% |
| 47 | GPT 5.4xhigh· Epoch AI Benchmarking Hub data | 49% |
| 48 | Claude Haiku 5.5max· Epoch AI Benchmarking Hub data | 46.3% |
| 49 | DeepSeek V4.1 Flash 1017max· Epoch AI Benchmarking Hub data | 46.3% |
| 50 | Muse Spark 1.3max· Epoch AI Benchmarking Hub data | 46.3% |
Showing 1–50 of 104 results.
Data credits: BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links. · Epoch AI Benchmarking Hub, CC BY 4.0 (licence) · Data by LLM Stats