RuneBench Skill - Smithing
Games · 2026-02-26
RuneBench 30-minute Smithing skill task. Agents play RuneScape through the SDK and are scored by the best real-game Smithing XP per minute achieved in any 15-second window.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-5.6 Sol | 375 |
| Claude Fable 5 | 338 |
| Gemini 3.5 Flash | 313 |
| Grok 4.5 | 313 |
| GPT-5.5 | 188 |
| GPT-5.4 | 119 |
| GPT-5.6 Luna | 113 |
| Opus 4.8 | 106 |
| Opus 4.7 | 81.0 |
| Sonnet 5 | 75.0 |
| Opus 4.6 | 63.0 |
| Gemini 3 Pro Preview | 62.0 |
| GLM-5.2 | 62.0 |
| GPT-5.4 Mini | 62.0 |
| Opus 4.5 | 56.0 |
| Kimi K2.7 Code | 50.0 |
| Kimi K2.6 | 44.0 |
| Gemini 3.1 Pro Preview | 34.0 |
| Qwen3.7-Max | 19.0 |
| Gemini 3 Flash Preview | 12.0 |
| Sonnet 4.6 | 6.0 |
| Qwen3 Max (rolling alias) | 3.0 |
| Haiku 4.5 | 0.0 |
| Sonnet 4.5 | 0.0 |
| DeepSeek-V4-Pro | 0.0 |
| GLM-5 | 0.0 |
| GPT-5.3-Codex | 0.0 |
| GPT-5.4 Nano | 0.0 |
| Grok 4.3 | 0.0 |
| Kimi K2.5 | 0.0 |
| Qwen3.5 35B A3B | 0.0 |