Google finally releases Gemini 4 Argon: After months sitting

Google launched Gemini 4 Argon on October 13, 2026, after months sitting on the sidelines of the frontier AI race. The Verge's report by Robert Hart, a London-based reporter and Senior Tarbell Fellow who previously covered health, science, and tech for Forbes, was published at 4:00 PM UTC. Google says the long-awaited model competes with the industry's best and delivers "frontier reasoning and improved performance across complex workflows," while cutting hedging, repetition, and the cliches that plague AI chatbots.

This piece is for Pro and Ultra subscribers in the Gemini app and AI mode in Search, for Ultra subscribers on Antigravity, and for developers comparing Argon with rival frontier models. It covers what Google shipped, where Argon is already running inside the company, the introductory token price, and the unsettled status of Plus subscribers and free users.

Google finally: what actually changed

Gemini 4 Argon is Google's first major frontier release since Gemini 3 in late 2025. In the months between, models from Anthropic and OpenAI left Google trailing far behind its rivals. Seismic internal shakeups raised further doubts about the company's ability, or its desire, to reclaim its position and compete at the frontier of AI. The October 13, 2026 release is Gemini 4 Argon, the long-awaited model Google says can compete with the best the industry has to offer after those months on the sidelines.

Google claims Argon outperforms GPT-6 Astra, Claude Fable 5.1, and Claude Opus 5.5 on benchmarks in coding, cybersecurity, science, finance, and law. Google says Argon delivers frontier reasoning and improved performance across complex workflows, and that it cuts hedging, repetition, and the cliches that continue to plague AI chatbots. The launch fills in blanks left when Google first teased Argon a few weeks ago. A limited release went first to a select set of "trusted cyber defenders," who tested the model for potential risks, in line with other recent frontier releases.

Google finally: how it works

Google says it is already putting Argon to use internally. The model is finding and patching cybersecurity vulnerabilities. It is halving the computing power needed to power YouTube's comments section. It is debugging the drone physics simulator for Wing, Alphabet's drone delivery company. Google says the model delivers frontier reasoning and improved performance across complex workflows, while cutting the hedging, repetition, and cliches that continue to plague AI chatbots. The company says Argon can compete with the best the industry has to offer.

Higher-tier Pro and Ultra subscribers get Argon in the Gemini app and in AI mode in Search. Ultra subscribers also get it on Antigravity, Google's agentic development platform. The public rollout on October 13, 2026, follows a limited release to a select set of "trusted cyber defenders" who tested Argon for potential risks, in line with other recent frontier model releases. Plus subscribers are not named in that rollout. Free users are not named either, and it is not clear when, or if, Argon will reach them.

Google finally: why it matters now

Google is pricing Argon very competitively. The introductory price is $2 per million input tokens and $10 per million output tokens, around a fifth the cost of OpenAI's GPT-6 Astra and Anthropic's Fable 5 models. Those costs double at the end of November, when the introductory period finishes. Even after the doubling, Argon will still be significantly cheaper than its frontier rivals on a per-token basis. A lower price per token does not necessarily mean lower cost in practice, because models can use different numbers of tokens for the same tasks.

Argon landed just a day after Meta announced its own frontier model, Muse Wonder Full. That sequence is a reminder of how quickly the balance of power can shift in AI, and that the frontier is far from a two-horse race. Argon is Google's first major frontier release since Gemini 3 in late 2025. In the months after Gemini 3, Anthropic and OpenAI fielded models that left Google trailing, and seismic internal shakeups raised doubts about Google's ability, or its desire, to compete at the frontier.

Google finally: who is affected

Argon rolls out on October 13, 2026, to higher-tier Pro and Ultra subscribers in the Gemini app and in AI mode in Search. Ultra subscribers also receive it on Antigravity, Google's agentic development platform. Plus subscribers and free users have no stated date. It is not clear when, or if, Argon will be available to them. The introductory price on the model is $2 per million input tokens and $10 per million output tokens, around a fifth the cost of GPT-6 Astra.

Google's own teams are already in the user group. Argon is finding and patching cybersecurity vulnerabilities, halving the computing power needed to power YouTube's comments section, and debugging the drone physics simulator for Wing, Alphabet's drone delivery company. A select set of "trusted cyber defenders" received a limited release before the public rollout, to test the model for potential risks. The models Google names as the ones Argon beats are GPT-6 Astra, Claude Fable 5.1, and Claude Opus 5.5. Meta's Muse Wonder Full arrived the day before.

Google finally: what to watch

At the end of November the introductory period finishes and Argon's costs double. The launch price is $2 per million input tokens and $10 per million output tokens, around a fifth the cost of GPT-6 Astra and Anthropic's Fable 5 models. After the doubling, Argon will still be significantly cheaper per token than those frontier rivals. Models can use different numbers of tokens for the same tasks, so the per-token gap does not necessarily mean a lower bill in practice.

Plus subscribers and free users still have no date after the October 13, 2026 rollout to Pro and Ultra in the Gemini app and AI mode in Search, and to Ultra on Antigravity. Google claims Argon outperforms GPT-6 Astra, Claude Fable 5.1, and Claude Opus 5.5 in coding, cybersecurity, science, finance, and law. Internally, the model is finding and patching vulnerabilities, halving the compute for YouTube comments, and debugging Wing's drone physics simulator. Meta announced Muse Wonder Full a day earlier.