Google ships Gemini 4 Argon, a long-task coding model, and it lands in GitHub Copilot

Google released Gemini 4 Argon on October 13, 2026, and positioned it squarely at long, multi-step professional work, with coding at the top of the list. On the same day it became generally available in the Gemini API and in GitHub Copilot, which means many people learning to build with AI can try it this week without changing tools.

What Google is claiming

In its launch post, Google describes Argon as a frontier model for multi-step work in coding, finance, legal, cybersecurity and science. The coding evidence it offers is mostly from its own engineering teams rather than public leaderboards:

• In internal use with Google's Antigravity coding tool, agentic code submissions rose by more than 35% and code rollbacks fell by nearly a third.
• A refactor of a roughly 6-million-line legacy folder inside Chrome's 40-million-line codebase took weeks instead of months, according to Google.
• Google also cites a 0.7% attack success rate on the Gray Swan indirect prompt-injection benchmark, which matters for agents that read untrusted web pages and files.

These are vendor-reported numbers measured on Google's own work. Treat them as a reason to try the model, not as proof it will beat your current one on your codebase.

The spec that stands out: output length

The headline developer change is how much the model can write in one go. Google calls it a "1M unified input-output token limit." The model page lists the API model ID as gemini-4-argon, a 1,048,576-token context window, and 65,536 output tokens by default, rising to the full 1,048,576 with long decoding. It supports function calling, code execution, structured outputs, and thinking levels (low, medium, high). Computer use is still marked as preview.

The Gemini API changelog lists the model as generally available from October 13.

Pricing has an expiry date

Introductory API pricing is $2 per million input tokens and $10 per million output tokens, with cached input 95% off. Google says that after the introductory period ends on November 30, 2026, the price doubles to $4 input and $20 output per million tokens. If you are benchmarking cost this month, budget for the later price.

Where you can use it

• GitHub Copilot: The GitHub changelog says Argon is generally available to Copilot Pro+, Max, Business and Enterprise users in the model picker across VS Code, Visual Studio, Copilot CLI, the Copilot coding agent, JetBrains IDEs, Xcode and Eclipse. It is billed at the provider's list price, and the rollout is gradual. Business and Enterprise admins control access through the Copilot model policy.
• Gemini app: Google AI Pro and Ultra subscribers get it in the Gemini app, and Antigravity's agentic coding interface is now built into the Gemini macOS app.
• Developers and enterprises: It is available through the Gemini API in AI Studio, Gemini Enterprise and Antigravity.

What it means for you

Each new frontier model adds another option to the model picker. The useful move is not to switch on launch day but to run Argon against a handful of real tasks from your own project, next to the model you already use, and compare both results and cost. Our evergreen guide this week walks through exactly how to do that.