FormalRewardBench - Pointwise: leaderboard
Pointwise accuracy (%) on FormalRewardBench (2026), a benchmark of reward models for Lean 4 theorem proving: the model scores a verified proof and a flawed proof of the same theorem independently, and an item counts when the correct one scores higher. Flawed proofs come from five error strategies such as minimal edits, forced mistakes and injected Python; higher is better.
Metric: Pointwise accuracy (%): the model scores the correct and the incorrect Lean proof of each item independently, and the item counts when the correct proof scores higher; temperature 1.0, top-p 0.75; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 26 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.5 | 70.1 | #160 |
| 2 | Claude Sonnet 4.5 | 62 | #261 |
| 3 | Con-J-Qwen2-7B | 52.8 | #329 |
| 4 | Gemini 2.5 Flash | 50.9 | #463 |
| 5 | Claude Sonnet 4 | 49.8 | #410 |
| 6 | GPT-5.2 | 48.9 | #185 |
| 7 | Selene-1 | 46.8 | |
| 8 | GPT-4o | 44 | #751 |
| 9 | GPT-4.1 | 44 | #505 |
| 10 | CompassJudger-1-7B-Instruct | 43.6 | |
| 11 | GPT-5.1 | 41.8 | #240 |
| 12 | LMUnit-qwen2.5-72B | 41.2 | #524 |
| 13 | CompassJudger-1-14B-Instruct | 40.3 | |
| 14 | Qwen 2.5 72B Instruct | 39.8 | #1117 |
| 15 | Skywork-Critic-Llama-3.1-70B | 39.2 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=formalrewardbench-pointwise · How It Works · Data refreshed daily, snapshot 2026-11-21.