Summary
Google’s newly released Gemini Argon 4 is put head-to-head with Anthropic’s Opus 5.5 in this hands-on test. Creator Jack Roberts gives both models identical prompts across seven levels of difficulty, tracking output quality, build time, and token cost for each task.
The challenges include designing a luxury perfume website with AI-generated video and imagery, producing a 15-second animated habit-tracker product intro, and building a browser-based Photoshop-style photo editor using the popular “reverse engineer anything” GitHub repository. In the website round, Opus 5.5 finished in roughly 30 minutes while Gemini took about 18 minutes longer, though Gemini was noticeably cheaper per token. The photo editor comparison shows Gemini offering more adjustment tools, while Claude’s version had better filters and layers but some usability quirks.
The video also covers how to access Argon 4 through the Gemini app, Google’s Antigravity IDE, or the API, and notes Google’s claim that the model leads on 12 of 18 general benchmarks. Viewers interested in choosing between frontier models for real-world coding and creative builds will find practical, side-by-side evidence rather than benchmark charts alone.
📺 Source: Jack Roberts · Published October 16, 2026
🏷️ Format: Comparison
