RuneBench Skill - Hitpoints
Games · 2026-02-26
RuneBench 30-minute Hitpoints skill task. Agents play RuneScape through the SDK and are scored by the best real-game Hitpoints XP per minute achieved in any 15-second window.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Fable 5 | 148 |
| Grok 4.5 | 139 |
| GPT-5.6 Sol | 126 |
| Opus 4.8 | 100 |
| DeepSeek-V4-Pro | 100 |
| Sonnet 5 | 69.0 |
| GPT-5.5 | 69.0 |
| GPT-5.6 Luna | 64.0 |
| Qwen3.7-Max | 60.0 |
| Kimi K2.6 | 48.0 |
| Kimi K2.7 Code | 45.0 |
| Gemini 3.5 Flash | 40.0 |
| Grok 4.3 | 38.0 |
| Opus 4.6 | 36.0 |
| Opus 4.7 | 35.0 |
| Haiku 4.5 | 32.0 |
| Sonnet 4.6 | 29.0 |
| Qwen3 Max (rolling alias) | 28.0 |
| GLM-5 | 27.0 |
| Gemini 3 Flash Preview | 26.0 |
| Gemini 3 Pro Preview | 26.0 |
| Sonnet 4.5 | 25.0 |
| GPT-5.4 | 23.0 |
| GPT-5.3-Codex | 21.0 |
| Gemini 3.1 Pro Preview | 17.0 |
| GPT-5.4 Mini | 16.0 |
| Qwen3.5 35B A3B | 14.0 |
| GPT-5.4 Nano | 13.0 |
| Opus 4.5 | 12.0 |
| GLM-5.2 | 11.0 |
| Kimi K2.5 | 5.0 |