Gemini 4 Argon in GitHub Copilot: $2 per million input tokens until Dec 1

Paid Copilot plans got Google's newest model on its release day. They pay for it at the provider's list price, not out of their request quota.

Where to turn it on. The model picker in VS Code, Copilot CLI and the Copilot app, plus five more clients: Visual Studio, JetBrains IDEs, Xcode, Eclipse and the coding agent. The rollout is gradual, so the model may not be in your list yet.

Billed by the token. Until Dec 1, a million input tokens cost $2, cached input $0.10 and output $10. The charge comes out of AI credits under usage-based billing, where one credit equals one cent. The model has no premium request multiplier at all.

On Business and Enterprise, an admin grants access through the model policy in Copilot settings. New models are on there by default until someone turns them off.

In GitHub's early tests, Argon solves more tasks than earlier Gemini models on every coding benchmark. GitHub sees the biggest gain in Copilot CLI but gives no numbers.

Google's own measurements paint a more mixed picture. On DeepSWE v1.1, Argon scores 77.9% against 74.2% for Claude Opus 5.5, while on Terminal-Bench 4.0 it scores 57.4% against 66.4%, so Opus 5.5 leads there.

The model has a single window: up to 1 million tokens of input and output via the continuation protocol. The protocol is meant for especially heavy tasks, not everyday work.

The promo price ends on Dec 1: input goes up to $4 per million tokens and output to $20.