# Vals ProofBench (ProofBench)

> Vals AI automated theorem-proving benchmark.

Canonical page: https://benchlm.ai/benchmarks/valsproofbench

- Category: [Mathematics](/math)
- Last updated: September 28, 2026

## About ProofBench

- Year: 2026
- Tasks: Automated theorem proving
- Format: Accuracy score
- Difficulty: Formal proof reasoning
- Paper: [ProofBench](https://www.vals.ai/benchmarks/proof_bench)

BenchLM mirrors Vals ProofBench as a display-only math and proof benchmark.

ProofBench is currently displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.

## Leaderboard (44 models)

| Rank | Model | Creator | Score |
|------|-------|---------|-------|
| 1 | [Claude Sonnet 5.5](/models/claude-sonnet-5-5) | Anthropic | 100.00% |
| 2 | [Claude Opus 5.5](/models/claude-opus-5-5) | Anthropic | 100.00% |
| 3 | [Claude Fable 5.1](/models/claude-fable-5-1) | Anthropic | 100.00% |
| 4 | [Alephprover](https://www.vals.ai/models/logicalintelligence_alephprover) | logicalintelligence | 100.00% |
| 5 | [GPT-6 Astra](/models/gpt-6-astra) | OpenAI | 99.00% |
| 6 | [Claude Opus 5](/models/claude-opus-5) | Anthropic | 99.00% |
| 7 | [Claude Fable 5](/models/claude-fable) | Anthropic | 95.00% |
| 8 | [Kimi K3](/models/kimi-k3) | Moonshot AI | 87.00% |
| 9 | [Aristotle](https://www.vals.ai/models/aristotle_aristotle) | Aristotle | 86.00% |
| 10 | [GPT-6 Sol](/models/gpt-6-sol) | OpenAI | 83.00% |
| 11 | [GPT-5.6 Sol](/models/gpt-5-6-sol) | OpenAI | 83.00% |
| 12 | [Claude Sonnet 5](/models/claude-sonnet-5) | Anthropic | 77.00% |
| 13 | [Hy4 preview](/models/hy4-preview) | Tencent | 75.00% |
| 14 | [GPT-5.6 Terra](/models/gpt-5-6-terra) | OpenAI | 74.00% |
| 15 | [MiMo-V2.6-Pro](/models/mimo-v2-6-pro) | Xiaomi | 70.00% |
| 16 | [GPT-6 Luna](/models/gpt-6-luna) | OpenAI | 64.00% |
| 17 | [MiMo-V2.6-Flash](/models/mimo-v2-6-flash) | Xiaomi | 63.00% |
| 18 | [GPT-5.6 Luna](/models/gpt-5-6-luna) | OpenAI | 60.00% |
| 19 | [Gemini 3.7 Flash](/models/gemini-3-7-flash) | Google | 58.00% |
| 20 | [Muse Spark 1.3 Max](https://www.vals.ai/models/meta_muse_spark_1_3_max) | Meta | 58.00% |
| 21 | [Qwen3.8 Max](/models/qwen3-8-max) | Alibaba | 58.00% |
| 22 | [DeepSeek V4 Flash 0731](/models/deepseek-v4-flash-0731) | DeepSeek | 56.00% |
| 23 | [Muse Spark 1.3](/models/muse-spark-1-3) | Meta | 55.00% |
| 24 | [DeepSeek V4.1 Flash](/models/deepseek-v4-1-flash) | DeepSeek | 54.00% |
| 25 | [Grok 4.6](/models/grok-4-6) | xAI | 51.00% |
| 26 | [DeepSeek V4 Pro 0813](/models/deepseek-v4-pro-0813) | DeepSeek | 50.00% |
| 27 | [GLM-5.3](/models/glm-5-3) | Z.AI | 49.00% |
| 28 | [Gemini 3.8 Flash](/models/gemini-3-8-flash) | Google | 48.00% |
| 29 | [Muse Spark 1.2](/models/muse-spark-1-2) | Meta | 43.00% |
| 30 | [Gemini 3.5 Flash](/models/gemini-3-5-flash) | Google | 31.00% |
| 31 | [Grok 4.5](/models/grok-4-5) | xAI | 31.00% |
| 32 | [Grok 4.7](/models/grok-4-7) | xAI | 26.00% |
| 33 | [Gemini 3.1 Pro Preview](https://www.vals.ai/models/google_gemini-3.1-pro-preview) | Google | 26.00% |
| 34 | [MiMo-V2.5-Pro](/models/mimo-v2-5-pro) | Xiaomi | 22.00% |
| 35 | [GLM-5.3-Flash](/models/glm-5-3-flash) | Z.AI | 21.00% |
| 36 | [MiniMax M3](/models/minimax-m3) | MiniMax | 18.00% |
| 37 | [DeepSeek V4 Pro 0813](/models/deepseek-v4-pro-0813) | DeepSeek | 16.00% |
| 38 | [MiMo-V2.5](/models/mimo-v2-5) | Xiaomi | 16.00% |
| 39 | [Qwen3.8-27B](/models/qwen3-8-27b) | Alibaba | 16.00% |
| 40 | [Mistral Medium 3.5](https://www.vals.ai/models/mistralai_mistral-medium-3.5) | Mistral | 9.00% |
| 41 | [Inkling-Small](/models/inkling-small) | Thinking Machines Lab | 6.00% |
| 42 | [Mercury 2.5](/models/mercury-2-5) | Inception | 3.00% |
| 43 | [Inkling](/models/inkling) | Thinking Machines Lab | 0.00% |
| 44 | [Nemotron Lightning 3p5 30b A3b](https://www.vals.ai/models/fireworks_nemotron-lightning-3p5-30b-a3b) | Fireworks | 0.00% |

## FAQ

### What does ProofBench measure?

Vals AI automated theorem-proving benchmark.

### Which model leads the published ProofBench snapshot?

Claude Sonnet 5.5 currently leads the published ProofBench snapshot with a score of 100.00%.

### How many models are evaluated on ProofBench?

The September 28, 2026 contains 44 AI models.

### Does ProofBench affect BenchLM's overall score?

Not directly. ProofBench is still displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.