Google launches Gemini 4 Argon, claiming it rivals top AI models

Google released Gemini 4 Argon on Tuesday, its first major frontier AI model since Gemini 3 launched in late 2025. The company says Argon delivers frontier reasoning, a term for a model’s ability to work through complex, multi-step problems, and improved performance across complex workflows, while cutting down on the hedging, repetition, and cliches that have dogged AI chatbots.

Google claims Argon beats rival models including OpenAI’s GPT-6 Astra, Anthropic’s Claude Fable 5.1, and Claude Opus 5.5 across a range of benchmarks, among them coding, cybersecurity, science, finance, and law. The company has not published independent verification of these results, and the claims come from Google’s own announcement.

Google says it is already using Argon inside the company. The model has been deployed to find and patch cybersecurity vulnerabilities, to cut the computing power needed to run YouTube’s comments section in half, and to debug the drone physics simulator used by Wing, Alphabet’s drone delivery business.

Argon is rolling out today to Google’s higher-tier Pro and Ultra subscribers in the Gemini app and in AI mode in Search, as well as to Ultra subscribers using Antigravity, Google’s agentic development platform for building AI-driven coding tools. Google has not said when, or whether, Argon will reach Plus subscribers or free users.

Pricing undercuts rivals, for now

Google is pricing Argon aggressively: $2 per million input tokens and $10 per million output tokens at launch, roughly a fifth of what OpenAI charges for GPT-6 Astra and what Anthropic charges for its Fable 5 models, according to Google. That introductory pricing holds until the end of November, when costs double. Even then, Google says Argon will remain cheaper than its frontier rivals on a per-token basis. That comparison does not necessarily mean lower costs in practice, since different models can use different numbers of tokens to complete the same task.

Before today’s public release, Google gave a limited group it calls “trusted cyber defenders” early access to Argon to test for potential risks, a step that has become standard practice ahead of frontier model launches at other AI labs.

The release follows a stretch in which Google trailed OpenAI and Anthropic on new model launches, a gap that widened alongside major internal reorganizations that raised questions about the company’s ability to keep pace at the frontier of AI development. Argon also arrives one day after Meta announced its own frontier model, called Muse Wonder Full, underscoring how crowded the top of the AI market has become.

What happens next will depend on how Argon performs once developers and independent testers put its benchmark claims to the test, and on whether Google extends access beyond Pro and Ultra subscribers before the introductory pricing expires at the end of November.