GitHub deprecated selected models in GitHub Copilot on October 2, 2026, affecting Copilot Chat, inline edits, ask and agent modes, and code completions. The change is important for teams that have model-specific workflows or policies because model availability can change independently of application code.
What changed
GitHub retired a group of older Copilot model options and provides newer alternatives. The broader Copilot lifecycle notice also sets October 19 as the next major deprecation date for additional models, including GPT-5.5, GPT-5.4 variants, Gemini 3.7 Flash, GPT-5 mini, and Grok 4.5.
Why model lifecycle management matters
AI applications increasingly depend on model identifiers, capabilities, context limits, pricing, and tool behavior. A model retirement can therefore change latency, cost, output quality, or tool behavior even when the surrounding application has not changed.
Enterprise implications
Enterprise administrators should inventory which Copilot models teams rely on and test suggested replacements before a retirement date. Organizations using explicit model policies may need to enable the replacement models or adjust internal standards.
Migration testing
Model replacement should be treated like a software dependency upgrade. Run representative coding tasks, review outputs, compare failure modes, and check whether custom prompts or agent instructions behave differently with the new model.
Practical takeaway
Teams using Copilot should maintain a small model-compatibility test suite and avoid assuming that a model identifier will remain available indefinitely. The easiest migration is one that has already been tested before the retirement date.
Source: GitHub Changelog — Selected models in GitHub Copilot deprecated