Google CEO Sundar Pichai Announces Global Launch of Gemini 4 Argon
Google CEO Sundar Pichai announced that Gemini 4 Argon is now globally available for developers and enterprises on the paid tier of the Gemini API and platform, Google Antigravity Ultra subscribers, and Ultra and Pro subscribers in the Gemini app and search. He included a blog link for complete availability, pricing, and demonstration information.
The model was first disclosed on September 30, initially launched through the Fairwind project to trusted cybersecurity defenders, showcasing cutting-edge performance in complex workflows, cybersecurity defense, and software engineering. The output token limit has been increased to 1 million, significantly expanding from the previous 64K. The company claims it ranks highly on the Vals Index, which measures economic-related tasks in finance, programming, law, and taxation.
The entry price for the API is $2 per million input tokens and $10 per million output tokens, with a 95% discount on cached inputs, valid until November 30, after which prices will rise to $4 and $20 respectively. Argon supports a unified context window of 1 million tokens, long decoding, and adjustable thinking levels. Google AI Pro subscribers can use it in the Gemini app and search, but it is limited to Ultra subscribers in Antigravity and AI Studio.
Over 100,000 Google employees have tested the model in the past month. Logan Kilpatrick stated that the model excels in coding persistence and long outputs, and it has been available to cybersecurity defenders and early customers for about two weeks. Free users and some lower-tier subscribers do not have access to Argon.
The announcement was publicly made by Pichai. Paid developers and enterprises gain access through the API, while Ultra and Pro subscribers benefit from the model's capabilities in the app and search; free and lower-tier users face capability stratification. Google is expanding the cutting-edge model from limited defenders to paid and subscription tiers through a phased rollout.
Source: Public Information
ABAB AI Insight
Sundar Pichai first disclosed Gemini 4 Argon on September 30 on X, with a limited launch to trusted defenders through the Fairwind project. On October 13, he announced its global launch, covering paid API, Antigravity Ultra, and Pro and Ultra subscribers in the Gemini app and search. The model's output limit has been increased to 1 million tokens, with an entry price of $2 per million input tokens and $10 per million output tokens until November 30.
Access is expanding from limited cybersecurity defenders to paid API and subscription tiers. Google is stratifying model capabilities: developers and enterprises access it through the paid API, while consumers use it in the app and search through Pro and Ultra subscriptions, with Antigravity available only to Ultra. Over 100,000 employees have tested it internally, and the company claims it performs well on the Vals Index and long-cycle software engineering benchmarks.
Similar to other frontier labs that initially open their strongest models to paid or enterprise tiers, Google is transitioning Argon from a limited Fairwind launch to paid and subscription distribution. It is currently in the phase of expanding the model from limited security testing to commercial availability, with pricing and subscription tiers determining the actual reach.
This essentially represents a transfer of pricing power. Free and lower-tier users cannot access the strongest model, while paid APIs and high subscription tiers control access. This change occurs because long outputs and complex reasoning consume higher computational resources, and the company is expanding cutting-edge capabilities from limited testing to users who can pay, while retaining a basic model for free tiers.
ABAB News · Cognitive Law
Subscription and API payment tiers determine the actual user reach of the frontier model
The global launch after limited security testing expands capabilities from defenders to commercial tiers
Entry pricing and subsequent adjustments bind long-term usage costs to model capabilities