RewardBench Ties — leaderboard
Metric: Score (%). Source: huggingface.co. 197 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4.1 | 85.42 |
| 2 | Claude Opus 4 (20250514) | 83.75 |
| 3 | Gemini 2.5 Flash (Preview 04-17) | 83.41 |
| 4 | Skywork-Reward-Gemma-2-27B-v0.2 | 81.82 |
| 5 | Gemini 2.5 Pro | 81.1 |
| 6 | Gemini 2.5 Flash | 80.9 |
| 7 | Claude Sonnet 4 (20250514) | 79.39 |
| 8 | GPT-4o (2024-08-06) | 78.19 |
| 9 | GPT-4.1 Mini | 74 |
| 10 | Gemini 2.5 Pro (Preview 05-06) | 69.73 |
| 11 | GPT-4o Mini (2024-07-18) | 69.62 |
| 12 | Claude 3.5 Sonnet (20240620) | 67.4 |
| 13 | Claude 3.7 Sonnet (20250219) | 67.23 |
| 14 | ArmoRM-Llama3-8B-v0.1 | 66.29 |
| 15 | Claude 3 Opus (20240229) | 56.01 |
Interactive version: theaggregate.ai/benchmark?slug=rewardbench-ties · How the rankings work · Data refreshed daily, snapshot 2026-07-22.