AICodeKing tests Gemini 3.8 Flash and Muse Spark 1.3 on eight tasks spanning browser simulations, 3D scenes, SVG generation, mathematics and an end-to-end fine-tuning workflow. Gemini scores 65 out of 80, or 81.25%, while Muse Spark scores 57 out of 80, or 71.25%.
Gemini's strongest results include a working bow-and-arrow game, a polished contact-lens case and a successful autonomous fine-tuning task. Muse Spark improves on elevator logic and sets the reviewer's best result for a dual-time-zone wristwatch, but regresses sharply on the panda SVG and folding-table tests that had distinguished version 1.2.
The practical warning applies to both models: each can overwrite whole files when a smaller edit was requested. The reviewer recommends small commits, careful diff inspection and workflow-level planning and verification. Product promotion and subscription requests are omitted.
Watch on YouTube
