RuneBench Skill - Magic
Games · 2026-02-26
RuneBench 30-minute Magic skill task. Agents play RuneScape through the SDK and are scored by the best real-game Magic XP per minute achieved in any 15-second window.
Top models (higher is better)
| Model | Score |
|---|---|
| Gemini 4 Argon | 2061 |
| GPT-6 Astra | 1820 |
| GPT-6.1 Sol | 1788 |
| GPT-6 Sol | 675 |
| Claude Opus 5.5 | 604 |
| Grok 4.6 | 421 |
| Claude Sonnet 5.5 | 320 |
| Grok 4.5 | 237 |
| GLM-5.2 | 203 |
| GPT-5.6 Sol | 179 |
| Claude Opus 5 | 177 |
| GPT-5.6 Terra | 156 |
| DeepSeek V4.1 Flash | 147 |
| GPT-6 Luna | 145 |
| Opus 4.8 | 130 |
| Opus 4.6 | 100 |
| Claude Fable 5.1 | 99.0 |
| Muse Wonder | 96.0 |
| GPT-5.3-Codex | 90.0 |
| Grok 4.7 | 90.0 |
| Claude Haiku 5.5 | 85.0 |
| Muse Spark 1.1 | 81.0 |
| GPT-5.5 | 80.0 |
| Sonnet 5 | 79.0 |
| Muse Spark 1.3 | 77.0 |
| Gemini 3.8 Flash | 76.0 |
| Gemini 3.5 Flash | 74.0 |
| Claude Fable 5 | 72.0 |
| Gemini 3.6 Flash | 72.0 |
| Opus 4.7 | 62.0 |
| Gemini 3 Flash Preview | 53.0 |
| GLM-5 | 48.0 |
| Sonnet 4.5 | 46.0 |
| GPT-5.4 | 46.0 |
| DeepSeek-V4-Pro | 43.0 |
| GPT-5.4 Mini | 39.0 |
| Gemini 3.7 Flash | 33.0 |
| GPT-5.6 Luna | 33.0 |
| Mistral Large 4 Preview | 32.0 |
| Gemini 3.1 Pro Preview | 31.0 |