Google released Gemini 4 Argon on October 13, claiming broad benchmark gains and offering initial access to paid subscribers and API users.
Google launched Gemini 4 Argon on October 13, introducing a new frontier model for its Gemini app, Search’s AI mode and developer platform Antigravity. The initial rollout is limited to Google Pro and Ultra subscribers in the Gemini app and AI mode, and Ultra subscribers on Antigravity. Google has not said whether or when free users or Plus subscribers will get access.
The company says Argon improves performance on complex tasks and reduces repetitive or overly cautious responses. Google also claims the model surpasses GPT-6 Astra, Claude Fable 5.1 and Claude Opus 5.5 across benchmarks covering areas including coding, cybersecurity, science, finance and law. Those are Google’s reported results; the article does not provide benchmark scores or independent validation.
Google is also putting Argon to work internally, according to the company. Reported uses include detecting and fixing security vulnerabilities, reducing the computing power used for YouTube comments by half, and debugging the drone physics simulator for Wing, Alphabet’s drone-delivery business.
For API users, the introductory price is $2 per million input tokens and $10 per million output tokens. Google plans to double those rates when the introductory period ends at the close of November. The lower per-token price does not necessarily mean every task will cost less: different models may consume different numbers of tokens to complete the same work.
The launch follows a limited test with selected “trusted cyber defenders,” part of a process to assess potential risks before broader availability. Argon is Google’s first major frontier-model release since Gemini 3 in late 2025. Its launch adds a new contender to a field where Google says it can now compete across several demanding domains, while access remains restricted to specific paid tiers.