RuneBench Skill - Firemaking
Games · 2026-02-26
RuneBench 30-minute Firemaking skill task. Agents play RuneScape through the SDK and are scored by the best real-game Firemaking XP per minute achieved in any 15-second window.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Fable 5 | 3189 |
| Grok 4.5 | 2632 |
| Gemini 3.5 Flash | 2227 |
| GPT-5.6 Sol | 1417 |
| DeepSeek-V4-Pro | 1125 |
| GPT-5.5 | 810 |
| Opus 4.7 | 600 |
| GLM-5.2 | 540 |
| Kimi K2.7 Code | 506 |
| GPT-5.6 Luna | 425 |
| Opus 4.8 | 370 |
| GPT-5.3-Codex | 360 |
| Opus 4.5 | 280 |
| Opus 4.6 | 280 |
| Sonnet 4.6 | 280 |
| Sonnet 4.5 | 260 |
| Sonnet 5 | 260 |
| Grok 4.3 | 260 |
| GPT-5.4 Mini | 240 |
| Kimi K2.6 | 240 |
| GPT-5.4 | 220 |
| GPT-5.4 Nano | 200 |
| Gemini 3.1 Pro Preview | 190 |
| Gemini 3 Flash Preview | 180 |
| Gemini 3 Pro Preview | 160 |
| Kimi K2.5 | 140 |
| Haiku 4.5 | 40.0 |
| Qwen3.7-Max | 40.0 |
| GLM-5 | 20.0 |
| Qwen3.5 35B A3B | 20.0 |
| Qwen3 Max (rolling alias) | 0.0 |