JamSet/JamBench: leaderboard
Current AI-driven game development has made substantial progress in asset generation, gameplay design, and web-based game coding, yet project-level code engineering on professional game engines remains largely unexplored due to the absence of large-scale datasets and deterministic evaluation methods. We present JamSet and JamBench, the first project-level game code framework dataset and benchmark
Metric: Task 1a SCS (self-reported). Source: benchmarklist.com. Saturation forecast: Around August 2027. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.4 | 46 | #79 |
| 2 | Claude Opus 4.6 | 42 | #63 |
| 3 | Kimi K2.5 | 39 | #144 |
| 4 | Gemini 3.1 Pro (Preview) | 37 | #58 |
| 5 | GLM-5 | 37 | #141 |
| 6 | Qwen 3.5 397B A17B | 35 | #146 |
| 7 | DeepSeek V4 Pro | 34 | #102 |
| 8 | Qwen 3.5 27B | 27 | #187 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=jamset-jambench · How It Works · Data refreshed daily, snapshot 2026-10-20.