GPT-5.3-Codex Is Now GitHub Copilot's First Long-Term Support Model
Enterprise software teams have long had a complicated relationship with AI model upgrades — every new release means re-running security reviews, updating compliance documentation, and re-validating behavior across internal toolchains. GitHub and OpenAI are taking a direct swing at that friction with the launch of a formal long-term support program for GitHub Copilot, and GPT-5.3-Codex is its inaugural model.
GPT-5.3-Codex launched on February 5, 2026, and under the new LTS designation it will remain available to enterprise customers through February 4, 2027 — a full twelve months of stability. More significantly, it's slated to replace GPT-4.1 as the default base model for all Copilot Business and Enterprise organizations by May 17, 2026. GitHub points to the model's "significantly high code survival rate" among enterprise customers as the data-driven rationale behind the selection, meaning code written by GPT-5.3-Codex is actually being kept and shipped, not discarded or heavily rewritten by human reviewers.
The move is a pointed response to a very real adoption blocker in larger organizations. Risk-averse engineering and security teams have been hesitant to build deep integrations with AI coding tools precisely because model churn can unexpectedly invalidate the policies and procedures built around them. By committing to a stable Codex model for a year, GitHub is essentially issuing an enterprise SLA for AI-generated code — and positioning Copilot as the safer, more auditable bet against rivals who offer raw capability without the governance scaffolding.