RP-Bench - Objective: leaderboard
RP-Bench objective slice measures role-playing model quality on automatically checked criteria for instruction following and character adherence.
Metric: Objective Score. Source: huggingface.co. Saturation forecast: Estimated already saturated. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4.1 | 90 | #248 |
| 2 | DeepSeek V3.2 | 89 | #205 |
| 3 | GLM-4.7 | 88.9 | #192 |
| 4 | Claude Opus 4.6 | 88 | #63 |
| 5 | Gemini 2.5 Flash | 86.8 | #245 |
| 6 | Qwen 3.5 Flash | 86.6 | #232 |
| 7 | Claude Sonnet 4.5 | 86.3 | #142 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=rp-bench-objective · How It Works · Data refreshed daily, snapshot 2026-10-20.