GitHub published two Copilot model-retirement notices on May 7, 2026. Claude Sonnet 4 was deprecated across Copilot experiences on May 6, with Claude Sonnet 4.6 listed as the suggested alternative. GPT-4.1 is scheduled to be deprecated across Copilot on June 1, 2026, with GPT-5.5 listed as the suggested alternative.
Both notices apply broadly across Copilot Chat, inline edits, ask and agent modes, and code completions.
Why this matters
Copilot is no longer a static coding assistant with a predictable model list. It is a managed multi-model platform where older models can retire across several developer surfaces at once.
That is good for keeping users on stronger models, but it is a risk for teams that pin workflows, docs, evals, or developer training to specific model names.
Buyer take
If you administer Copilot, keep a model retirement calendar. Check whether GPT-5.5 and Claude Sonnet 4.6 are enabled in organization policies before the GPT-4.1 deadline.
If you are an individual user, stop treating GPT-4.1 results as a long-term baseline for Copilot comparisons. Re-test prompts, agent workflows, and code-review behavior on GPT-5.5 or Sonnet 4.6 before making buying decisions.
What is still unclear
GitHub did not publish migration impact data, such as how many completions or agent sessions still rely on GPT-4.1 or Claude Sonnet 4.
Sources
Primary and corroborating references used for this news item.