Google Gemini 4 Argon Guide

Gemini 4 Argon is Google’s new frontier AI model, announced on September 30, 2026. Google positions it as its frontier model for real-world coding, enterprise knowledge work, and cyber defense. It is a new model, not a fine-tune of an older one.

Google built it for:

• Long tasks with many steps
• Complex software engineering
• Legal, finance, and other professional tasks
• Charts, long videos, and many documents at the same time
• Creative writing

Argon is built to finish long jobs, not just field quick questions.

September 30

Trusted cyber defenders and internal Google teams got it first

Next

Paid API customers and Google AI Ultra subscribers came next

October 13

Google AI Pro and Ultra subscribers got access in the Gemini app

Google staged the release for safety. Free Gemini users did not get access at launch.

Context window

Roughly 1 million tokens

Output

Up to 1 million tokens, up from 64,000 previously

Inputs

Text, images, video, audio, and PDF files

Output type

Text

• Answers that are more direct, with fewer caveats and stock phrases
• More progress updates during long jobs
• A tasks mode toggle that lets Gemini keep working while you are away
• Connections to more than 40 apps, such as Instacart, OpenTable, Adobe, and Shopify
• Simpler handoff to Google Antigravity’s agentic coding, now inside the Gemini macOS app

Model ID

gemini-4-argon

Where

The Gemini API paid tier in Google AI Studio and Android Studio, plus Google Cloud for enterprises

Launch price

$2 per million input tokens, $10 per million output tokens

Later price

$4 input and $20 output once the introductory period ends

Caching

95% off the price of cached input tokens

Google reports that Argon:

• Set a new DeepSWE v1.1 record for software engineering, at 77.9%
• Tops the Vals Index for professional tasks
• Places first on Zapier’s AutomationBench
• Shares first place on the CWE-bench cybersecurity benchmark
• Tops a benchmark for resisting indirect prompt injection

These results come from Google itself. Test Argon on your own work.

• Large coding projects and migrations
• Research that spans many long documents
• Financial and legal analysis, reviewed by experts
• Insights drawn from long videos and charts
• Multi-step tasks that run across your connected apps

• State the goal and define what done looks like
• Share the files and links it should work from
• Set the rules, budget, and deadline
• Ask for progress updates along the way
• Name what needs your approval

Review these five contracts. Put every renewal date, notice period, and price increase in one table, flag unusual terms, and cite the page for each item. Do not contact anyone.

• Approve anything that buys, books, sends, or shares on your behalf
• Connect only the apps a task actually needs
• Review your Gemini Apps activity and data settings
• Keep passwords and sensitive personal data out
• Have experts check legal, medical, and financial outputs

• Access varies by plan and country
• Long tasks and huge outputs eat more of your limits or API budget
• It can still get things wrong, so verify important facts
• Most benchmarks are self-reported so far

Goal: [what you want done]. Context: [who you are, files, and links]. Rules: [limits, budget, deadline, tone]. Output: [format and length]. Work step by step, send me progress updates, cite your sources, and ask before any action that sends, buys, books, or shares.