On May 17, 2026, GitHub announced that GPT-5.3-Codex is now the base model for all Copilot Business and Copilot Enterprise organizations, replacing GPT-4.1. At first this looks like a model switch. For enterprise AI coding governance, the more important themes are stability, internal approval, and long-term support.
The base model is what Copilot uses when an organization has not yet approved other models through its internal review process. That matters because large companies often require security, legal, data protection, and technical review before a model can be used broadly. The base model therefore affects the everyday experience of many developers.
GitHub also says GPT-5.3-Codex is its first long-term support model in partnership with OpenAI. It launched on February 5, 2026 and will remain available through February 4, 2027 for Copilot Business and Enterprise users. That 12-month window is important because frequent model changes create work for security reviews, regression testing, documentation, and internal training.
Cost and policy are part of the update too. GPT-5.3-Codex carries a 1x premium request multiplier. GPT-4.1 remains force-enabled at a 0x multiplier for now, but GitHub says it will deprecate alongside usage-based billing on June 1, 2026. Enterprises therefore need to manage quality, cost, and model policy together.
This update shows AI coding tools moving into a more mature procurement and governance phase. Early adopters often asked which model felt best. Enterprises increasingly need to ask which model is stable for review, how long it will be supported, whether it passes internal controls, how workflows change after the switch, and how cost will be forecast.
For development teams, GPT-5.3-Codex as the base model means more agentic coding capability will be present by default in enterprise environments. A base model designed for longer tasks, tool use, and code context can make Copilot agent mode, review assistance, and automated fixes feel more natural in daily work.
Still, model upgrades do not remove the need for local acceptance standards. Teams should keep checking test coverage, security review, code ownership, pull request rules, and human approval points. As AI coding enters a steadier phase, the advantage will shift from trying the newest model first to managing model governance, cost, and engineering quality together.
