On October 13, Google opened Gemini 4 Argon, which it calls its frontier model for complex workflows, to developers and to the Gemini app. The model had been announced on September 30, but at the time only select cybersecurity partners had access.

What Google released

According to the developer post, Argon targets long-horizon software engineering, enterprise knowledge work and cybersecurity defense, with a focus on deep reasoning and structural precision in tasks such as code migrations and frontier research. It is available on the paid tier of the Gemini API, in Google AI Studio and in Android Studio. Google AI Ultra subscribers can try it in AI Studio and in Google Antigravity, the company's agentic coding tool.

In the Gemini app, the model is available to AI Pro and Ultra subscribers. Google says answers are more direct, with less hedging and repetition, and that the model shares progress more often on long tasks. The app gains a toggle between chat and tasks, where Gemini keeps working even when the user is away from the computer, across Google apps, and Antigravity's agentic coding interface is now in the Gemini macOS app. Connected apps now exceed 40, including Instacart, OpenTable, Adobe and Shopify.

The limited September 30 debut

According to Khaleej Times, Google introduced Argon on Wednesday, September 30, as the anchor of the Gemini 4 generation and larger than its previous "Pro" models. The company described it as its "most performant model yet built for complex workloads." It was given to cybersecurity partners at the time, and Google was taking part in the US government's voluntary pre-release model access process, with no timeline for public release.

Performance: what is a claim

Google's release, per Khaleej Times, says Argon beat OpenAI's Astra and Anthropic's Opus on several industry benchmarks but trailed both on some, including two of the four coding benchmarks in the release. The figures are published by the company itself. The official launch posts for developers and for the app give no price, context window or benchmark results, so the site does not report them.

Context

Khaleej Times reports that Google dropped Gemini 3.5 Pro, which CEO Sundar Pichai had said would arrive in June, and that the company overhauled DeepMind, with Demis Hassabis stepping aside and several Gemini leaders leaving. The outlet says the delays left Google behind Anthropic and OpenAI. The October 13 release is the first time the general public, at least paying users, can use the model.

What we still don't know

  • API pricing and context window size, which are not in the official posts.
  • Independent test results for the model.
  • When a version for free-tier users will arrive.