GameCraft-Bench: Can Agents Build Playable Games End-to-End in a Real Game Engine?

GameCraft-Bench is a newly released benchmark designed to evaluate the capability of AI agents to construct playable games end-to-end within a real game engine. It includes assessments of various large models such as Opus-4.7 and GPT-5.5, while also raising interest in the performance of medium-sized models like Qwen3.6-27B and Gemma-4-31B in comparison to larger counterparts. This benchmark is significant for practitioners as it provides insights into the potential of different model sizes for game development tasks, which can influence model selection and optimization strategies in AI-driven game design.