ProofBench leaderboard
- Claude Sonnet 5.5 leads
- 100.0%
- models measured
- 74
- at the top
- Tied
- median
- 40.5%
Leader on ProofBench · read October 21
On October 21, 2026, 74 of our catalog's models had a ProofBench result, and Anthropic Claude Fable 5.1 led with 100.0%.
Epoch AI collects its results; its benchmarking hub describes the tasks and links the authors.
Role in the index
We collect this benchmark and show it as supporting evidence. It is not part of the Vibe Coding Index, because it measures something other than coding and the index is about coding. Read the methodology
| # | Model | Result | Cost per test | Time per proof | Evidence | Coding index | Output $/Mtok |
|---|---|---|---|---|---|---|---|
| 1 | Claude Fable 5.1Anthropic | 100.0% | $2.7035 | 7m 58s | 89.1 | $50.00 | |
| 2 | Claude Opus 5.5Anthropic | 100.0% | $0.9616 | 5m 51s | 92.2 | $20.00 | |
| 3 | Claude Sonnet 5.5Anthropic | 100.0% | $0.5583 | 5m 48s | 88.9 | $10.00 | |
| 4 | Claude Opus 5Anthropic | 99.0% | $2.1062 | 9m 51s | 81.8 | $25.00 | |
| 5 | Gemini 4 ArgonGoogle | 99.0% | $0.783 | 6m 20s | 80.5 | $10.00 | |
| 6 | GPT-6 AstraOpenAI | 99.0% | $1.7124 | 5m 08s | 88.7 | $50.00 | |
| 7 | GPT-6.1 SolOpenAI | 99.0% | $0.2331 | 4m 45s | 88.7 | $10.00 | |
| 8 | Claude Fable 5Anthropic | 95.0% | $5.4515 | 13m 43s | 82.1 | $50.00 | |
| 9 | Claude Haiku 5.5Anthropic | 89.0% | $0.1739 | 6m 58s | 75.9 | $0.50 | |
| 10 | Kimi K3Moonshot AI | 87.0% | $1.6468 | 34m 31s | 61.8 | $11.936 | |
| 11 | Muse WonderMeta | 84.0% | $1.3133 | 15m 29s | 64.5 | $20.00 | |
| 12 | GPT-5.6 SolOpenAI | 83.0% | $1.4383 | 8m 40s | 76.7 | $20.00 | |
| 13 | GPT-6 SolOpenAI | 83.0% | $0.4763 | 5m 05s | 82.8 | $10.00 | |
| 14 | Claude Sonnet 5Anthropic | 77.0% | $1.3737 | 14m 58s | 63.5 | $10.00 | |
| 15 | Hy4 previewTencent | 75.0% | $0.3357 | 22m 54s | 61.6 | $2.501 | |
| 16 | GPT-5.6 TerraOpenAI | 74.0% | $1.2742 | 12m 57s | 69.0 | $12.00 | |
| 17 | MiMo-V2.6-ProXiaomi | 70.0% | $0.2454 | 1h 02m | 72.6 | $0.87 | |
| 18 | Claude Opus 4.8Anthropic | 69.0% | $2.2931 | 19m 26s | 66.0 | $25.00 | |
| 19 | GPT-6 LunaOpenAI | 64.0% | $0.0416 | 7m 31s | 71.4 | $0.50 | |
| 20 | MiMo-V2.6-FlashXiaomi | 63.0% | $0.1233 | 1h 03m | 61.9 | $0.28 | |
| 21 | GPT-5.6 LunaOpenAI | 60.0% | $0.1233 | 8m 34s | 61.5 | $1.20 | |
| 22 | Gemini 3.7 FlashGoogle | 58.0% | $0.5585 | 7m 04s | 69.9 | $3.75 | |
| 23 | Muse Spark 1.3Meta | 58.0% | $0.479 | 9m 29s | 79.7 | $4.25 | |
| 24 | Qwen3.8 Max (0902)Qwen | 58.0% | $1.504 | 31m 08s | 63.5 | $6.00 | |
| 25 | DeepSeek V4 Flash 0731DeepSeek | 56.0% | $0.0366 | 15m 34s | Not benchmarked | $1.32 | |
| 26 | GPT-5.4OpenAI | 56.0% | $3.202 | 45m 52s | 37.7 | $15.00 | |
| 27 | Claude Opus 4.7Anthropic | 54.0% | $1.8078 | 11m 20s | 51.1 | $25.00 | |
| 28 | DeepSeek V4.1 FlashDeepSeek | 54.0% | $0.1264 | 9m 07s | 77.2 | $0.525 | |
| 29 | Grok 4.6xAI | 51.0% | $0.759 | 14m 22s | 76.4 | $6.00 | |
| 30 | Claude Opus 4.6Anthropic | 50.0% | $2.7036 | 1h 06m | 66.9 | $25.00 | |
| 31 | DeepSeek V4 Pro 0813DeepSeek | 50.0% | $0.0622 | 12m 12s | 54.6 | $1.98 | |
| 32 | GPT-5.5OpenAI | 50.0% | $1.6818 | 49m 13s | 67.8 | $30.00 | |
| 33 | GLM 5.3Z.ai | 49.0% | $2.0815 | 42m 23s | 73.1 | $1.40 | |
| 34 | Gemini 3.8 FlashGoogle | 48.0% | $0.6029 | 6m 09s | 70.5 | $4.125 | |
| 35 | Claude Sonnet 4.6Anthropic | 45.0% | $2.1621 | 1h 31m | 57.8 | $15.00 | |
| 36 | Muse Spark 1.2Meta | 43.0% | $0.442 | 7m 49s | 65.1 | $4.25 | |
| 37 | Step 5 PreviewStepFun | 42.0% | $0.7146 | 38m 34s | 71.4 | $2.70 | |
| 38 | Muse Spark 1.1Meta | 39.0% | $0.5197 | 17m 37s | 48.9 | $4.25 | |
| 39 | Claude Opus 4.5Anthropic | 36.0% | $1.267 | 1h 03m | 66.8 | $25.00 | |
| 40 | Gemini 3.6 FlashGoogle | 36.0% | $0.5221 | 5m 47s | 48.3 | $3.75 | |
| 41 | GLM 5.2Z.ai | 35.0% | $0.00 | 0.0 s | 57.7 | $7.00 | |
| 42 | Gemini 3.5 FlashGoogle | 31.0% | $0.5699 | 6m 34s | 42.8 | $9.00 | |
| 43 | Grok 4.5xAI | 31.0% | $0.574 | 11m 09s | 75.5 | $6.00 | |
| 44 | Gemini 3.1 Pro PreviewGoogle | 26.0% | $0.7843 | 14m 49s | 35.3 | $12.00 | |
| 45 | Qwen3.7 MaxQwen | 26.0% | $2.4621 | 12m 51s | 45.1 | $4.425 | |
| 46 | Grok 4.7xAI | 26.0% | $0.788 | 12m 47s | 75.0 | $6.00 | |
| 47 | GLM 5.1Z.ai | 22.2% | $0.5846 | 1h 19m | 44.0 | $4.40 | |
| 48 | MiMo-V2.5-ProXiaomi | 22.0% | $0.0899 | 19m 37s | 40.0 | $0.609 | |
| 49 | GPT-5.4 MiniOpenAI | 21.0% | $0.2494 | 23m 31s | 28.2 | $4.50 | |
| 50 | GLM 5.3 FlashZ.ai | 21.0% | $0.2242 | 1h 11m | 71.0 | $0.50 | |
| 51 | Claude Sonnet 4.5Anthropic | 19.0% | $0.6534 | 3h 25m | 59.1 | $15.00 | |
| 52 | MiniMax M3MiniMax | 18.0% | $0.4192 | 11m 59s | 57.8 | $1.20 | |
| 53 | GPT-5OpenAI | 18.0% | $0.8524 | 1h 16m | 72.2 | $10.00 | |
| 54 | DeepSeek V4 Pro 0423DeepSeek | 16.0% | $0.0336 | 7m 03s | 35.3 | $1.9105 | |
| 55 | Kimi K2.6Moonshot AI | 16.0% | $0.2958 | 23m 07s | 44.2 | $3.41 | |
| 56 | Qwen3.8 27BQwen | 16.0% | $2.2988 | 24m 14s | 73.6 | $1.35 | |
| 57 | MiMo-V2.5Xiaomi | 16.0% | $0.0571 | 20m 56s | 34.0 | $0.238 | |
| 58 | Gemini 3 Flash PreviewGoogle | 15.0% | $0.3273 | 19m 02s | 75.8 | $3.00 | |
| 59 | GPT-5.2OpenAI | 15.0% | $1.0597 | 24m 54s | 58.9 | $14.00 | |
| 60 | Gemini 3.5 Flash LiteGoogle | 13.0% | $0.1328 | 4m 51s | 33.6 | $2.50 | |
| 61 | GPT-5 NanoOpenAI | 12.0% | $0.0492 | 1h 02m | 34.8 | $0.40 | |
| 62 | Grok 4.3xAI | 11.0% | $0.2441 | 27m 23s | 22.5 | $2.50 | |
| 63 | Mistral Large 4Mistral AI | 10.0% | $0.6424 | 20m 59s | 55.3 | $4.18 | |
| 64 | Mistral Medium 3.5Mistral AI | 9.0% | $1.302 | 7m 59s | 9.2 | $7.50 | |
| 65 | GPT-5 MiniOpenAI | 9.0% | $0.2202 | 55m 50s | 59.8 | $2.00 | |
| 66 | GPT-5.1-Codex-MaxOpenAI | 9.0% | $0.5274 | 53m 04s | Not benchmarked | $10.00 | |
| 67 | T Inkling SmallThinking Machines | 6.0% | $0.0754 | 8m 27s | 29.9 | $1.20 | |
| 68 | GLM 4.7Z.ai | 6.0% | $0.144 | 56m 05s | 33.6 | $1.98 | |
| 69 | GPT-5.4 NanoOpenAI | 5.0% | $0.0159 | 8m 45s | Not benchmarked | $1.25 | |
| 70 | MiniMax M2.5MiniMax | 4.0% | $0.0575 | 1h 15m | 59.4 | $0.95 | |
| 71 | Mercury 2.5Inception | 3.0% | $0.0441 | 3m 22s | Not benchmarked | $0.15 | |
| 72 | MiniMax M2.7MiniMax | 3.0% | $0.0429 | 27m 14s | 28.4 | $1.20 | |
| 73 | Nemotron 3 UltraNVIDIA | 2.0% | Not published | 42m 32s | Not benchmarked | $2.40 | |
| 74 | T InklingThinking Machines | 0.0% | $0.2014 | 7m 16s | 49.7 | $4.05 |
Only models in our catalog are listed, and each appears once with its best published run. 197 rows were read from the source on October 21, 2026, and rows that match no catalog model are left out. Cost per test is the spend on one proof, in US dollars, as Epoch AI records it beside the result. Every other column is a figure the board publishes for the same run.