Google Releases Gemini 4 Argon Model to Pro and Ultra Subscribers

Google has released its Gemini 4 Argon model to Google AI Pro and Ultra subscribers. The company announced availability in the Gemini app on 13 October 2026. Argon is positioned as a frontier model for complex reasoning, software engineering, and enterprise tasks.

Google states that Argon answers more directly, hedges less, and repeats itself less often than earlier versions. It also shares progress updates during longer tasks. The model is already in use inside Google for coding assistance, research, and certain internal systems, including cybersecurity work and YouTube comment moderation. A toggle in the Gemini macOS app now allows users to switch between chat and an agentic coding interface powered by Antigravity.

Introductory pricing is set at $2 per million input tokens and $10 per million output tokens, with cached input tokens at a 95 percent discount. Google says the rate is lower than current offerings from OpenAI and Anthropic on a per-token basis. The introductory period is scheduled to end in late November, after which prices are expected to rise but remain below some competitors.

Argon scores highly on internal benchmarks for long-horizon software engineering and on economic-impact measures that weight finance, legal, and tax work by contribution to US GDP. The model is not yet available to free users or to lower-tier Plus subscribers. Google has described a phased rollout that includes voluntary pre-release access for selected cyber defenders.

The release follows months in which Google trailed some rivals in public frontier-model announcements. Argon is intended to close that gap for paid users who need stronger performance on multi-step workflows. The company has not published full independent benchmark comparisons against the latest versions of GPT or Claude models.

Developers and enterprises can access Argon through the Gemini app and, for Ultra subscribers, through the Antigravity platform. Google notes that the model connects to more than 40 external apps for task completion, ranging from grocery ordering to design tools. Users who select the tasks mode can assign work that continues when they are away from the computer.

The pricing and capability claims will be tested as independent evaluators run the model on public benchmarks and real-world workloads. For now, Argon is limited to higher-tier subscribers while Google gathers feedback and adjusts guardrails. A broader release date has not been set.