Summary
With the release of Gemini 4 and the launch of Meta’s Muse Wonder, the frontier model race now has four serious contenders. Peter Yang, who had early access to all but Muse, tests Gemini 4, Claude Opus 5.5, GPT-6 Astra, and Muse Wonder across five practical use cases: 3D models, games, videos, website design, and writing.
The first test asks each model to build Himeji Castle in Blender and render a 30-second cinematic at dawn during cherry blossom season, complete with music, entirely in code. Opus 5.5 comes out on top, with Gemini 4 a clear runner-up, Astra third, and Muse Wonder last. Meta’s Alex Wang has said the model was not trained on 3D models. A second test asks for a Star Fox-style rail shooter using StarCraft assets, where Opus is again the most polished and Gemini adds the most features.
Other tests include emotional short videos, where Muse Wonder surprisingly produces the most moving story, and a from-scratch redesign of the creator’s newsletter brand and homepage. The video offers a grounded look at where each model leads and where it falls short.
📺 Source: Peter Yang · Published October 13, 2026
🏷️ Format: Comparison
