GitHub Copilot model deprecation will take effect on October 19 2026, when a set of older generation models is removed from all Copilot experiences, including Chat, inline edits, ask, agent modes, and code completions. Teams that rely on these models for automated code generation, CI‑CD assistance, or custom tooling need to confirm that their configurations point to the suggested replacement models before the deadline, otherwise the calls will fall back to the default enabled models.
Models being retired and their suggested alternatives
- Gemini 3.7 Flash →
Gemini 3.8 Flash - GPT‑5.5 →
GPT‑5.6 Sol - GPT‑5.4 →
GPT‑5.6 Sol - GPT‑5.4 mini →
GPT‑5.6 Luna - GPT‑5 mini →
GPT‑5.6 Luna - Grok 4.5 →
Grok 4.6
Default model enablement behavior
For Copilot Enterprise and Business customers, the replacement models are automatically turned on under the global default setting. If an administrator has disabled the global default or explicitly turned off a specific model, the new model must be enabled manually via the model policies in the Copilot settings. Once enabled, the model appears in the Copilot Chat selector and is available to all supported experiences.
Implementation and operational implications
Practitioners should audit any automation that references the deprecated model identifiers—such as API calls, CI‑CD scripts, or IDE extensions—to ensure they either rely on the default selection or explicitly request the new model name. Because the deprecation does not require a removal step, calls that still reference the old identifiers will simply resolve to the enabled default, but this may lead to unexpected model behavior if the alternative differs in prompt handling or token limits.
From an operational standpoint, the change introduces a brief window where both old and new models could be active, depending on policy settings. Teams should verify that monitoring dashboards, usage quotas, and cost reports reflect the new model names to avoid surprises in billing or quota exhaustion.
Security and compliance considerations
The source does not call the retired models insecure, but any shift in model version can affect data handling policies, especially for organizations that enforce model‑specific data retention or audit rules. Review any compliance documentation that references model identifiers and update it to the new names. If custom model policies were used to restrict data flow, confirm that the new models inherit the same policy constraints.
Related CloudNinjas coverage: AI engineering.
What This Means For Practitioners
Action items: (1) inventory all places where a Copilot model name is hard‑coded; (2) replace deprecated identifiers with the suggested alternatives or rely on the default enablement; (3) if global defaults are disabled, enable the new models via the Copilot settings UI; (4) update monitoring, cost, and compliance records to reflect the new model names; and (5) test critical workflows after the switch to confirm expected output quality before the October 19 deadline.
