Godot Benchmark 2: Opus 5 > Sol > Terra (ziva.sh)

The latest Godot Benchmark has evaluated three AI models—Claude Opus 5, GPT 5.6 Sol, and GPT 5.6 Terra—in their ability to create a 3D vampire survivor game from an AI-generated game specification. Opus 5 emerged as the top performer, delivering a playable game experience with intricate design elements, while Sol provided a rougher but functional alternative at a significantly lower cost and faster turnaround. Conversely, Terra failed to produce a playable game, with attacks that never connected, highlighting its limitations. This benchmark is significant for the AI/ML community as it underscores the varying capabilities of different models in practical applications like game development. The findings indicate that while Opus 5 provides a robust output, its high cost and longer creation time may deter developers, suggesting a trade-off between quality and efficiency. With Sol's lower price point and acceptable results, it presents an attractive option for those looking for quicker gameplay prototyping. The evaluation emphasizes the importance of iterating on AI-generated content to refine usability and player interaction, essential factors in game design.