RuneBench Skill - Prayer
Games · 2026-02-26
RuneBench 30-minute Prayer skill task. Agents play RuneScape through the SDK and are scored by the best real-game Prayer XP per minute achieved in any 15-second window.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Opus 5.5 | 623 |
| Gemini 4 Argon | 563 |
| GPT-5.6 Terra | 563 |
| GPT-6.1 Sol | 563 |
| GPT-6 Astra | 563 |
| GPT-6 Sol | 563 |
| GPT-5.5 | 562 |
| GPT-5.6 Luna | 562 |
| Grok 4.6 | 562 |
| Gemini 3.7 Flash | 247 |
| DeepSeek V4.1 Flash | 223 |
| Grok 4.7 | 214 |
| GPT-5.6 Sol | 210 |
| Claude Fable 5.1 | 157 |
| Gemini 3.8 Flash | 137 |
| Claude Opus 5 | 99.0 |
| Claude Sonnet 5.5 | 81.0 |
| Opus 4.8 | 79.0 |
| Opus 4.7 | 63.0 |
| Sonnet 5 | 63.0 |
| Gemini 3.1 Pro Preview | 63.0 |
| GLM-5.2 | 63.0 |
| GLM-5.3 | 63.0 |
| GPT-5.4 | 63.0 |
| GPT-6 Luna | 63.0 |
| Muse Spark 1.1 | 63.0 |
| Muse Spark 1.2 | 63.0 |
| Grok 4.5 | 61.0 |
| GPT-5.3-Codex | 58.0 |
| Claude Fable 5 | 54.0 |
| DeepSeek-V4-Flash-0731 | 54.0 |
| Mistral Large 4 Preview | 43.0 |
| Gemini 3.5 Flash | 38.0 |
| Gemini 3.6 Flash | 37.0 |
| Gemini 3 Flash Preview | 23.0 |
| GPT-5.4 Mini | 22.0 |
| GPT-5.4 Nano | 22.0 |
| Muse Wonder | 22.0 |
| Gemini 3.5 Flash-Lite | 20.0 |
| Grok 4.3 | 20.0 |