Google Opens Gemini 4 Argon to Paying Users Worldwide, With $2 Intro Pricing That Doubles After November 30

Google has begun the wide rollout of Gemini 4 Argon, its first top-tier model since Gemini 3 arrived in late 2025, opening it to paying subscribers and developers worldwide on Wednesday after weeks in which access was restricted to a small set of vetted security partners.

AI Pro and Ultra subscribers can now use Argon in the Gemini app and in AI Mode in Search. Ultra subscribers additionally get it inside Antigravity, Google's agentic development platform, where an interface integration now surfaces the coding environment directly inside the macOS Gemini app. The company has not said when or whether the model will reach Plus subscribers or free users, who were recently pushed down to Flash-Lite models.

For developers, Argon carries introductory API pricing of $2 per million input tokens and $10 per million output tokens, with a 95 percent discount on cached input. The rates double to $4 and $20 after November 30 — a deadline that gives teams a limited window to evaluate workflows at launch economics. Reuters reported the introductory price is roughly a fifth of what OpenAI and Anthropic charge for comparable frontier models, though token counts vary by task.

The model's headline specification is a 1 million output token ceiling, up from the 64,000 figure Google disclosed when Argon was announced in September. That matters for long-horizon work: Google cites legal research, financial analysis, large codebase migrations and multi-step agent runs as the intended use cases. On the company's own numbers, which are self-reported and not independently verified, Argon scores 77.9 percent on DeepSWE v1.1 for long-horizon software engineering, 51.3 percent on Zapier's AutomationBench, 91.7 percent on the LVBench long-video benchmark, and a tied-first 68 percent on CWE-Bench v1 for vulnerability repair.

Google also offered rare detail on internal deployment, arguing thousands of Googlers already rely on the model. It says an Argon-assisted cleanup of redundant emoji rendering cut server load for YouTube's comment system by 45 percent, and that the model surfaced critical bugs in Wing's drone physics simulator that had sat in the codebase since 2018. On the engineering side, Google describes C-to-Rust migration work spanning from small base libraries up to the roughly 800,000-line Fuchsia Zircon kernel, and a rewrite of about 32,000 lines of SIMD code in the libgav1 video decoder that it says produced a Rust implementation 2.7 times faster than the existing port.

Teams using Argon in Antigravity increased agentic code submissions by more than 35 percent and cut code rollbacks by nearly a third, according to Google — again, company-reported figures. Early customers include Deloitte, LexisNexis, Ambience Healthcare and Isomorphic Labs. Google is simultaneously widening its Fairwind Program, which hands vetted cybersecurity defenders a guardrail-free build of the model, from a select group to hundreds of partners, and will give 10,000 scientists six months of free access to Antigravity's highest usage tier.

The rollout lands after a bruising stretch for Google's AI efforts. The company cancelled Gemini 3.5 Pro in September, months after Pichai had promised it for June, and restructured DeepMind after co-founder Demis Hassabis stepped aside and several Gemini leaders departed. Meta shipped its own frontier-tier Muse Wonder Full the day before. Argon's actual test now is whether a price that doubles in six weeks converts curious developers into committed ones — or sends them comparison shopping before December.