GPT-5.3-Codex Becomes the Copilot Business and Enterprise Base Model

GitHub started rolling GPT-5.3-Codex as the new base model for Copilot Business and Enterprise plans on May 17, 2026. It replaces GPT-4.1 as the default for the entire plan tier, and it is GitHub and OpenAI’s first long-term support model on Copilot: the LTS window guarantees the model stays selectable through 2027-02-04.

Individual seats (Copilot Pro, Pro+, Free) are not affected. The change only flips the default for Business and Enterprise.

What “base model” actually controls

Base model is the one Copilot uses when a request does not pin a specific model. Anywhere you wrote model: gpt-4.1 in a Copilot config, that is unchanged for now. Anywhere you let Copilot pick, the answer just changed from GPT-4.1 to GPT-5.3-Codex.

GPT-5.3-Codex has a 1x premium request multiplier, the same as GPT-4.1, so the per-request cost on Business and Enterprise SKUs does not move with this swap. Inline completions, Chat without a pinned model, and the cloud agent’s auto selection all flip together.

What changes for repos pinning the old default

Two places to scan before 2026-06-01. After that date, requests still pinned to gpt-4.1 start billing under the usage-based meter instead of being included.

# 1. Workflow files that pin a Copilot model
grep -RE "model:\s*gpt-4\.1" .github/ 2>/dev/null

# 2. Copilot agent and Chat custom instructions
grep -R "gpt-4.1" .copilot/ .github/copilot-instructions.md 2>/dev/null

If the project’s CI runs Copilot CLI or Copilot cloud agent tasks against a pinned GPT-4.1, there are two options: bump the pin to gpt-5.3-codex, or accept the line-item charge from June 1 forward. A YAML pin for the new default looks the same shape:

# .github/workflows/copilot-review.yml
- uses: github/copilot-action@v1
  with:
    model: gpt-5.3-codex
    effort: high

Why GitHub picked a Codex variant for the LTS slot

GPT-5.3-Codex is the code-tuned sibling in the GPT-5.3 family. GitHub’s stated metric in the rollout post was code survival rate, the share of accepted suggestions that are still in the file after subsequent edits and PR merges. The changelog reports a significantly higher rate among Business and Enterprise customers in the rollout cohort against GPT-4.1, which is the rationale for designating it the LTS base instead of the generalist GPT-5.3.

The LTS designation matters more than the model swap itself. GitHub deprecates models on a rolling basis with little notice. Claude Sonnet 4 was removed from every Copilot surface eleven days earlier with a two-paragraph changelog and no migration window. The Codex LTS commitment is GitHub’s first dated availability guarantee on a Copilot model, and the rest of the lineup does not get one.

GPT-4.1 access continues at no extra charge until 2026-06-01. After that, the meter starts.