Overview
Gemini 4 Argon is now an option for Google AI Pro and Ultra subscribers using the Gemini App. Existing subscribers can evaluate it on difficult work before considering a higher plan. For prospective subscribers, eligibility and workflow restrictions matter alongside capability claims.
This guide uses official information checked on October 14, 2026. Upgrade recommendations are editorial judgments; we have not independently tested Argon.
Two Launch Stages
| Date | Event | Meaning for an App user |
|---|---|---|
| September 30, 2026 | Model announced with initial Fairwind access | General App availability was not established. |
| October 13, 2026 | Argon opened to Pro and Ultra in the Gemini App | Eligible subscribers could start using it. |
The model announcement and App announcement describe two stages of one release.
The API release notes also record gemini-4-argon as generally available on October 13. Google's developer announcement identifies the paid API tier as a separate access route for developers.
What Changed
More room for sustained reasoning
Google raises the model output ceiling from 64K to 1M tokens for extended reasoning. This specification is separate from the App context window. It does not establish the amount of text available in a single App response.
Coding and professional problem-solving
Google reports 77.9% on DeepSWE v1.1, which evaluates long software engineering tasks, and describes document-heavy professional applications. These model results suggest workloads to evaluate; they do not guarantee identical results or faster responses in consumer chats.
More direct answers and clearer progress
Google describes less repetitive responses and better updates during extended tasks. Evaluate these product claims on a difficult prompt, including the time you spend checking and revising the result.
App workflows accompanying the rollout
The App announcement also introduces easier switching between chat and tasks, plus an Antigravity coding interface inside the Gemini macOS app. These accompanying App features should not all be attributed to Argon or assumed available on every device.
Compared With Previous Version
Gemini 3.8 Flash was already available to Pro and Ultra subscribers and supported coding and agent workflows. Argon does not introduce paid access or coding for the first time.
The model adds an emphasis on deeper, extended problem-solving and a stated increase in output capacity. Compare it with the model already handling your work: assess correctness, completion, and usage consumed. We have no independently measured App latency comparison.
Our recommendation: try Argon on a task your current model struggles to finish before making it your default. Routine prompts may still suit a faster model.
Availability & Access
Google's personal-account model table lists Argon for Pro and Ultra in chat and task modes. Neither the free tier nor Google AI Plus includes Argon or task mode. A paid Plus subscription is insufficient.
Eligible accounts can choose a model through the prompt box's selector and adjust thinking levels. Work and school access follows Workspace licensing, rather than this personal-account table.
Task mode has additional requirements
The task guide requires a personal account, age 18 or above, Pro or Ultra, Keep Activity enabled, and opt-in to Personal Intelligence. It currently excludes the European Economic Area, Nigeria, Switzerland, and the United Kingdom. Argon chat access alone does not establish task eligibility.
Gemini task mode is experimental, and Google recommends user supervision. Eligible users can access tasks on mobile, Mac, and web. A remote browser can continue after your device closes; local browser work requires your device and Chrome to remain available. Some steps pause for user input.
Mac requirements
Google's Mac requirements specify macOS 15 or later, Apple Silicon, at least 8 GB RAM, and an internet connection. Check these before choosing a workflow centered on the desktop app.
Usage & Subscription Decision
Both Pro and Ultra include Argon. Choose based on required capacity rather than assuming Ultra is necessary to unlock the model. Check the local subscription offer before paying; this page does not establish a universal monthly price.
Google's limits guide uses compute-based allowances, influenced by prompt complexity, model, and conversation length. Capacity refreshes every five hours until the weekly limit is reached. Higher plans offer more capacity, but this is not unlimited use or a fixed daily prompt count.
The guide lists a 1M-token App context window for Pro and Ultra. This is information the App can consider, separately from Argon's output ceiling. Check remaining capacity under Settings → Usage Limits.
Who Should Upgrade / Who Should Wait
- Existing Pro or Ultra subscribers: try difficult code or multi-document analysis before paying for a higher tier.
- Free or Plus users with recurring complex work: consider Pro if evaluating these workloads justifies the local price.
- Subscribers hitting limits: check which activities consume capacity before choosing a higher allowance.
- Users mainly asking everyday questions: keep the model that already completes them adequately.
- Users subscribing mainly for automation: establish task eligibility first; model access alone is insufficient.
These are workload-based recommendations, not comparative test results.
