Summary
Developer and YouTuber Theo shares his early-access testing of Gemini 4 Argon, Google’s first full-size Gemini release in about eight months, since Gemini 3.1 Pro in February. The video compares it with recent models including Claude Opus 5.5, GPT 6.1 Soul, GPT 6 Astra, and Muse Wonderful.
Theo covers how Argon is priced and where it ranks on third-party measures such as the Artificial Analysis intelligence index, noting that its score puts it above Astra and 6.1 Soul while its token usage affects real cost. He discusses Google’s published claims, including the top spot on the Vals Index, and points out inconsistencies in the launch materials. He also looks at what makes Gemini models distinctive: strong access to YouTube data and APIs, solid vision work, and a standout result on Skatebench. At the same time, he is candid about weak spots, including a confusing Google product lineup and Gemini’s underwhelming performance in Google’s own consumer services.
The second half moves to day-to-day coding, frontend quality, and his 3D game test, Fishlop, to see how Argon handles real work. Theo discloses his relationships with Google and with sponsor WorkOS, whose agent sign-up standard is also briefly discussed.
📺 Source: Theo – t3․gg · Published October 14, 2026
🏷️ Format: Review
