RP-Leaderboard: leaderboard
Roleplay quality judged by Claude Sonnet 4.5: realism, instruction-following, character depth, worldbuilding, dialogue and narrative immersion in extended roleplay chats.
Metric: RP Score (%). Source: psychosmiley.github.io. Saturation forecast: Around January 2027. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Kimi K2.5 | 88.9 | #144 |
| 2 | Gemini 3 Pro (Preview) | 87.1 | #66 |
| 3 | Grok 4.1 Fast | 85.8 | #217 |
| 4 | Claude Opus 4.5 | 66.5 | #83 |
| 5 | GPT-5.1 Codex | 61.8 | #159 |
| 6 | Nemotron 3 Ultra | 56.6 | #235 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=rp-leaderboard · How It Works · Data refreshed daily, snapshot 2026-10-20.