ProfBench — leaderboard
Metric: Overall Rubric Score (%). Source: huggingface.co. 98 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.7 (Thinking) | 61.3 |
| 2 | GPT-5.6 Sol (xHigh) | 60.4 |
| 3 | Claude Opus 4.6 (Thinking) | 58.8 |
| 4 | GPT-5.5 (xHigh) | 58 |
| 5 | GPT-5.3 Codex (xHigh) | 57.2 |
| 6 | GPT-5.4 (xHigh) | 56.8 |
| 7 | Claude Opus 4.8 (Thinking) | 56.8 |
| 8 | Claude Sonnet 4.6 (Thinking) | 55.6 |
| 9 | Gemini 3.1 Pro (Preview) | 55.2 |
| 10 | GPT-5.2 (xHigh) | 55.1 |
| 11 | Gemini 3.6 Flash | 55 |
| 12 | GPT-5.1 (High) | 54.9 |
| 13 | Gemini 3.5 Flash | 54.5 |
| 14 | Claude Opus 4.5 (Thinking) | 54.2 |
| 15 | O3 | 52.4 |
Interactive version: aibenchmarks.dev/benchmark?slug=profbench · How the rankings work · Data refreshed daily, snapshot 2026-07-22.