GitHub brings GPT-5.3-Codex long-term support to Copilot for enterprise stability

GitHub announced on March 18, 2026 that it is introducing a long-term-support model policy for Copilot Business and Copilot Enterprise, with GPT-5.3-Codex designated as the first LTS model. GitHub says LTS models will remain available for a full 12 months from launch so enterprise customers can complete internal security and safety reviews without worrying about rapid model turnover. GPT-5.3-Codex launched on February 5, 2026 and, under this policy, will stay available through February 4, 2027.

The change is not just about retention. GitHub also said GPT-5.3-Codex will become the newest base model for Copilot, replacing GPT-4.1. The base model is what an organization gets when it has not yet approved additional models through its internal governance process. GitHub plans to enable GPT-5.3-Codex automatically within 60 days of the announcement, with May 17, 2026 set as the date it becomes the base model for all Copilot Business and Copilot Enterprise organizations. The company also said GPT-5.3-Codex carries a 1x premium request unit multiplier, while GPT-4.1 will remain force-enabled at a 0x multiplier for the time being.

Why it matters

• The policy formalizes a reality of enterprise AI adoption: stability, reviewability, and predictable availability matter almost as much as frontier performance.
• GitHub is turning model governance into product structure by separating the base model from the broader approved-model set an organization may allow.
• The move lowers friction for enterprises that want better coding models but cannot constantly re-run internal approval cycles.

The announcement is specifically targeted at Business and Enterprise tiers rather than all Copilot users, but it still signals a broader market shift. As coding assistants become deeper parts of production workflows, buyers increasingly care about operational continuity, procurement, and compliance. GitHub’s LTS policy suggests the next phase of competition in AI coding tools will be shaped not just by model quality, but by how manageable those models are inside enterprise change-control systems.