RuneBench Skill - Ranged
Games · 2026-02-26
RuneBench 30-minute Ranged skill task. Agents play RuneScape through the SDK and are scored by the best real-game Ranged XP per minute achieved in any 15-second window.
Top models (higher is better)
| Model | Score |
|---|---|
| Sonnet 5 | 246 |
| Gemini 3.5 Flash | 240 |
| Opus 4.7 | 228 |
| Claude Fable 5 | 166 |
| GPT-5.6 Sol | 144 |
| Gemini 3 Flash Preview | 138 |
| GPT-5.5 | 134 |
| Grok 4.5 | 130 |
| Opus 4.8 | 128 |
| GPT-5.6 Luna | 122 |
| Gemini 3.1 Pro Preview | 118 |
| Opus 4.6 | 114 |
| GLM-5.2 | 104 |
| GPT-5.3-Codex | 84.0 |
| Grok 4.3 | 64.0 |
| Opus 4.5 | 62.0 |
| Sonnet 4.5 | 42.0 |
| GPT-5.4 Mini | 36.0 |
| Sonnet 4.6 | 34.0 |
| Qwen3.7-Max | 34.0 |
| Qwen3 Max (rolling alias) | 32.0 |
| GPT-5.4 | 24.0 |
| DeepSeek-V4-Pro | 22.0 |
| Haiku 4.5 | 20.0 |
| Gemini 3 Pro Preview | 19.0 |
| GLM-5 | 16.0 |
| Kimi K2.6 | 12.0 |
| Kimi K2.7 Code | 8.0 |
| Qwen3.5 35B A3B | 6.0 |
| GPT-5.4 Nano | 0.0 |
| Kimi K2.5 | 0.0 |