Google's Gemini 4 Argon pricing doubles on December 1

Google lists Gemini 4 Argon through the Gemini API, Google AI Studio and the Gemini platform, and describes it as globally available to developers and enterprises. The model's unified context limit is 1,048,576 tokens, shared between input and output. The API identifier is gemini-4-argon. During the introductory period that ends Nov. 30, 2026, the API rate is $2 per 1 million input tokens and $10 per 1 million output tokens. From Dec. 1, those rates move to $4 and $20. Cached input costs $0.10 per 1 million tokens in the introductory window, a 95% discount.

Access does not include the API Free tier. Paid API customers can call gemini-4-argon. Google AI Pro and Ultra subscribers have access in the Gemini app and Search AI Mode, and Ultra subscribers also get it in Antigravity. Google also describes a limited six-month program at no charge for eligible scientists and researchers who apply. Neither Shattered nor Tech Insider found Malaysian subscription prices in the Google materials they reviewed, and neither confirmed a Malaysia-specific launch.

Google exposes thinking levels of low, medium and high. The default synchronous API output cap is 65,536 tokens. Long decode mode lets a response reach the full 1,048,576-token limit. That is up from 64,000 tokens in the previous Gemini generation, according to Tech Insider's comparison. Because input and output share the same pool, a prompt that fills the window leaves little room for a reply.

Google's statements describe global availability rather than a named Malaysia launch. Shattered and Tech Insider both reported that local coverage framed the launch as a Malaysia launch, going beyond Google's published documentation.

The December 1 change doubles every price line. After the increase, output still costs five times as much as input, and cached input remains 5 percent of the fresh input rate. Shattered lists the cached rate as $0.20 from Dec. 1, while Tech Insider says Google has not stated the post-November cached rate. Tech Insider calculated that a request reusing 900,000 cached tokens with a 65,536-token output costs $0.75 in the introductory period.

Reasoning tokens are billed as output, so a high thinking level can raise the bill quickly. Shattered advised teams to set explicit output limits and spend alerts, noting that a single long-decode request near the ceiling costs about $10 at the introductory rate and about $20 from Dec. 1. Tech Insider's analysis suggested compute costs as the likely driver for excluding the free tier, but Google has not stated its reasoning. The phrase "paid plans only" overstates the situation because Google also runs a six-month free program for eligible scientists and researchers, which requires an application.