Benchmarks / Coding
GameDevBench
Uses first-party pass@1 across all 333 game-development tasks. Score: pass@1 across all 333 tasks. Collected by Jake Boggs' benchmark dashboard.
GameDevBench
Higher is better · best effort per model
full table below
| # | Model | Effort | Score |
|---|---|---|---|
| 1 | GPT-6.1 Sol | xhigh | 70.0% |
| 2 | GPT-6 Astra | high | 68.8% |
| 3 | Fable 5 | max | 67.3% |
| 4 | GPT-5.6 Sol | xhigh | 63.7% |
| 5 | Muse Spark 1.2 | high | 61.0% |
| 6 | Kimi K3 | max | 58.0% |
| 7 | Opus 4.8 | max | 55.9% |
| 8 | GPT-5.5 | xhigh | 54.7% |
| 9 | Gemini 3 Pro | high | 53.8% |
| 10 | GPT-5.4 | xhigh | 52.0% |
| 11 | Gemini 3 Flash | high | 46.8% |
| 12 | GPT-5.4 Mini | xhigh | 43.2% |
| 13 | GLM 5.2 | max | 38.4% |
| 14 | Sonnet 4.5 | high | 34.8% |
| 15 | Kimi K2.5 | max | 20.7% |
| 16 | Haiku 4.5 | xhigh | 18.6% |
| 17 | Qwen 3.5 397B | – | 5.4% |