Google DeepMind compared its two AI models on the same task: building a representation of the Eiffel Tower. Gemini 4 Argon succeeded on the first attempt for $0.23, while Anthropic’s Claude Opus 5.5 required two hours of work and cost $33.45, roughly 145 times more expensive. This difference is partially explained by pricing, with Argon at $2 per million input tokens versus $4 for Opus 5.5, but mainly by first-pass efficiency. On 18 comparative tests, Argon won 14 categories, particularly in finance and legal applications, while Opus 5.5 leads on independent aggregate indices and terminal-heavy benchmarks. Argon’s promotional pricing expires on November 30, 2026.
Source: Read the original article


