Gemini 4 Argon goes GA at frontier parity and lower cost

Artificial Analysis says that Gemini 4 Argon matches GPT-6 Astra on its Intelligence Index at 60% of the cost. That cost advantage depends on a temporary price cut, not on Argon using fewer tokens. In its general-availability post, Artificial Analysis gives Argon a Cost per Task of $1.99 "with discounted prices", about 40% below GPT-6 Astra (max). Cost per Task is what it costs to run the model on the index's test tasks, so it depends both on how many tokens the model produces and on the price of each token. Google DeepMind launched Argon at an introductory $2/$10 per million input/output tokens, with cached input 95% off. Its launch post says $4/$20 will apply after the introductory period ends on November 30, 2026.

Artificial Analysis's earlier analysis explains why the discount matters. It says Argon's cost efficiency "is driven by lower token prices, rather than reduced token use". Argon averaged 62k output tokens per task, against 27k for GPT-6 Astra (max). At standard pricing, it put Argon's Cost per Task at $3.98, about 1.2 times Astra's. The same analysis puts the discounted $1.99 at 2.7 times the cost of GPT-6.1 Sol (max), which scored one point lower on the index. That article is dated September 30. At the time, Argon was rolling out only to selected users, after Google first gave access to trusted testers, and Google had not yet confirmed when the discount would end. The October general-availability post repeats the $1.99 figure and adds the November 30 date.