Skip to main content

Vals ProofBench (ProofBench)

Vals AI automated theorem-proving benchmark.

How BenchLM shows ProofBench

BenchLM mirrors the public Vals AI ProofBench leaderboard captured from https://www.vals.ai/benchmarks/proof_bench and updated by Vals on May 16, 2026. The snapshot preserves overall scores, uncertainty, latency, cost-per-test metadata, and task-level scores where Vals publishes them.

ProofBench is display only on BenchLM. Vals proprietary or Vals-hosted aggregate views are useful context, but BenchLM does not use them as weighted ranking inputs or as a replacement for benchmark-native source records.

32 Vals rows1 task viewsprivate datasetTasks: OverallDisplay only

ProofBench score on ProofBench — May 16, 2026

BenchLM mirrors the published proofbench score view for ProofBench. Aristotle leads the public snapshot at 71.00% , followed by GPT-5.4 (56.00%) and Claude Opus 4.7 (54.00%). BenchLM does not use these results to rank models overall.

32 modelsExternal benchmark mirrorsCurrentDisplay onlyUpdated May 16, 2026

The published ProofBench snapshot is tightly clustered at the top: Aristotle sits at 71.00%, while the third row is only 17.00 points behind. The broader top-10 spread is 48.78 points, so the benchmark still separates strong models even when the leaders cluster.

32 models have been evaluated on ProofBench. The benchmark falls in the External benchmark mirrors category. BenchLM tracks this category separately from its weighted global scoring system, so these results are best compared on the dedicated Korean benchmark views. ProofBench is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About ProofBench

Year

2026

Tasks

Automated theorem proving

Format

Accuracy score

Difficulty

Formal proof reasoning

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

BenchLM freshness & provenance

Version

ProofBench 2026

Refresh cadence

Quarterly

Staleness state

Current

Question availability

Public benchmark set

CurrentDisplay only

BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.

ProofBench score table (32 models)

1
Aristotlearistotle/aristotle
71.00%
2
GPT-5.4openai/gpt-5.4-2026-03-05
56.00%
3
Claude Opus 4.7anthropic/claude-opus-4-7
54.00%
4
GPT-5.5openai/gpt-5.5
50.00%
5
Claude Opus 4.6 Thinkinganthropic/claude-opus-4-6-thinking
50.00%
6
Claude Sonnet 4.6anthropic/claude-sonnet-4-6
45.00%
7
Claude Opus 4.5 20251101 Thinkinganthropic/claude-opus-4-5-20251101-thinking
36.00%
8
Gemini 3.5 Flashgoogle/gemini-3.5-flash
29.00%
9
Gemini 3.1 Pro Previewgoogle/gemini-3.1-pro-preview
26.00%
10
GLM 5.1 Thinkingzai/glm-5.1-thinking
22.22%
11
GPT-5.4 Miniopenai/gpt-5.4-mini-2026-03-17
21.00%
12
Gemini 3 Pro Previewgoogle/gemini-3-pro-preview
20.00%
13
Claude Sonnet 4.5 20250929 Thinkinganthropic/claude-sonnet-4-5-20250929-thinking
19.00%
14
GPT-5openai/gpt-5-2025-08-07
18.00%
15
Muse Sparkmeta/muse_spark
17.00%
16
Kimi K2.6 Thinkingkimi/kimi-k2.6-thinking
16.00%
17
Gemini 3 Flash Previewgoogle/gemini-3-flash-preview
15.00%
18
GPT-5.2openai/gpt-5.2-2025-12-11
15.00%
19
Grok 4.20 0309 Reasoninggrok/grok-4.20-0309-reasoning
14.00%
20
GPT-5 Nanoopenai/gpt-5-nano-2025-08-07
12.00%
21
Grok 4.3grok/grok-4.3
11.00%
22
DeepSeek V4 Prodeepseek/deepseek-v4-pro
10.00%
23
GPT-5.1 Codex Maxopenai/gpt-5.1-codex-max
9.00%
24
GPT-5 Miniopenai/gpt-5-mini-2025-08-07
9.00%
25
Qwen3.6 27balibaba/qwen3.6-27b
8.00%
26
DeepSeek V3p2fireworks/deepseek-v3p2
8.00%
27
GLM 4.7zai/glm-4.7
6.00%
28
GPT-5.4 Nanoopenai/gpt-5.4-nano-2026-03-17
5.00%
29
Grok 4.1 Fast Reasoninggrok/grok-4-1-fast-reasoning
4.00%
30
DeepSeek V3p2 Thinkingfireworks/deepseek-v3p2-thinking
4.00%
31
MiniMax M2.5 Lightningminimax/MiniMax-M2.5-Lightning
4.00%
32
MiniMax M2.7minimax/MiniMax-M2.7
3.00%

FAQ

What does ProofBench measure?

Vals AI automated theorem-proving benchmark.

Which model leads the published ProofBench snapshot?

Aristotle currently leads the published ProofBench snapshot with 71.00% proofbench score. BenchLM shows this benchmark for display only and does not use it in overall rankings.

How many models are evaluated on ProofBench?

32 AI models are included in BenchLM's mirrored ProofBench snapshot, based on the public leaderboard captured on May 16, 2026.

Last updated: May 16, 2026 · mirrored from the public benchmark leaderboard

The AI models change fast. We track them for you.

For engineers, researchers, and the plain curious — a weekly brief on new models, ranking shifts, and pricing changes.

Free. No spam. Unsubscribe anytime.