Effective September 1 2026, GitHub retired a set of Copilot models across Chat, inline edits, ask/agent modes, and code completions. The change matters because any pipeline, IDE integration, or automation that references these models will silently fall back or fail unless the recommended replacements are selected and the appropriate policies are enabled.
Copilot model deprecation overview
The following models are no longer available after the deprecation date, with the suggested alternatives listed alongside:
Gemini 3.1 Pro→Gemini 3.7 FlashClaude Opus 4.5→Claude Opus 4.7,Claude Opus 4.8, orClaude Opus 5Claude Opus 4.6→Claude Opus 4.7,Claude Opus 4.8, orClaude Opus 5Claude Sonnet 4.5→Claude Sonnet 5Claude Sonnet 4.6→Claude Sonnet 5Raptor Mini→MAI‑Code‑1.1‑Flash
Note that Claude Sonnet 4.6 remains accessible to individual Copilot subscribers on annual plans, but it is not part of the enterprise offering.
Updating policies and settings
Enterprise administrators must verify that the replacement models are permitted in their organization’s model‑policy configuration. This involves:
- Opening the Copilot settings for the organization.
- Ensuring the desired alternative model is toggled on in the model‑policy list.
- Confirming visibility of the model in the Copilot Chat selector within VS Code or on github.com.
If a policy blocks the new model, the UI will not present it, and any automated calls that request the deprecated model will receive an error response.
Operational impact and considerations
Because GitHub does not automatically purge references to the retired models, existing workflow files, CI/CD scripts, or custom integrations will continue to reference them until they are edited. The immediate effect is a potential loss of functionality or reduced quality of suggestions if the fallback model is less capable. Practitioners should audit:
- GitHub Actions that invoke Copilot APIs with explicit model names.
- IDE extensions or internal tools that hard‑code a model identifier.
- Enterprise policy dashboards for any lingering deny rules on the new models.
There is no required cleanup step for the deprecated models; they simply become unavailable.
Related CloudNinjas coverage: AI engineering.
What This Means For Practitioners
Review all code‑generation and AI‑assisted automation for references to the listed models, replace them with the suggested alternatives, and verify that the organization’s model policies allow those alternatives. Test the updated configurations in a staging environment before rolling out to production to avoid unexpected suggestion gaps. Finally, keep an eye on GitHub’s documentation for any future model changes that could affect your pipelines.
