On May 17, GitHub switched the base model for all Copilot Business and Copilot Enterprise organizations to GPT-5.3-Codex, retiring GPT-4.1 as the default. Per GitHub’s changelog, the change is automatic for those plans and does not touch individual Pro, Pro+, or Free subscriptions.

A long-term-support model

GPT-5.3-Codex is the first model GitHub has designated long-term support, a label it says it created in partnership with OpenAI. It comes with a 12-month availability guarantee: the model launched on February 5, 2026 and is committed to stay available for Business and Enterprise through February 4, 2027. GitHub frames this as stability for organizations that need a fixed model to run security and safety reviews against, rather than re-certifying every new release. Until now, Copilot defaults rotated whenever a newer model shipped; the LTS guarantee freezes that churn for a year.

What it costs

GPT-5.3-Codex carries a 1× premium-request multiplier. GPT-4.1 stays force-enabled at a 0× multiplier for now, but it deprecates when usage-based billing begins on June 1, 2026. GitHub says GPT-5.3-Codex shows a notably high code-survival rate among enterprise customers — the share of generated code that stays in the codebase rather than being reverted.

Not the only model change

The same week, GitHub made Gemini 3.5 Flash generally available in Copilot on May 19 and updated the model list on the Copilot web client on May 20. Those changes add models to the picker for developers who go looking; the base-model switch is different in kind, because it changes the default for everyone who does not. For most enterprise developers, the base-model change is the one that happened without them touching a setting — taken up in you didn’t pick this model.