RewardBench Focus — leaderboard
Metric: Score (%). Source: huggingface.co. 197 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude 3.7 Sonnet (20250219) | 92.12 |
| 2 | Skywork-Reward-Gemma-2-27B-v0.2 | 91.72 |
| 3 | Claude 3.5 Sonnet (20240620) | 86.97 |
| 4 | Gemini 2.5 Flash (Preview 04-17) | 86.72 |
| 5 | Claude Opus 4 (20250514) | 86.16 |
| 6 | Gemini 2.5 Flash | 84.1 |
| 7 | Gemini 2.5 Pro (Preview 05-06) | 83.08 |
| 8 | Gemini 2.5 Pro | 80.5 |
| 9 | ArmoRM-Llama3-8B-v0.1 | 76.57 |
| 10 | Claude Sonnet 4 (20250514) | 75.96 |
| 11 | GPT-4o Mini (2024-07-18) | 74.14 |
| 12 | GPT-4.1 Mini | 73.54 |
| 13 | GPT-4.1 | 73.38 |
| 14 | GPT-4o (2024-08-06) | 72.93 |
| 15 | Gemini 1.5 Flash-8B | 67.47 |
Interactive version: theaggregate.ai/benchmark?slug=rewardbench-focus · How the rankings work · Data refreshed daily, snapshot 2026-07-22.