GitHub has deprecated selected models across GitHub Copilot, with an October 2, 2026 change confirmed in its changelog and a second group scheduled for October 19 in daily.dev’s coverage. The changes span Copilot Chat, inline edits, ask and agent modes, and code completions. For engineering managers, the practical work is to identify affected model choices, move workflows to supported alternatives and check whether organizational policies allow those alternatives.
GitHub’s October 2 notice confirms the first deprecation date and its scope. The specific model lists and replacement guidance below come from daily.dev’s summaries. Its October 19 report describes a scheduled change; the available evidence does not separately confirm that the later rollout was completed.
Which models are affected?
October 2: Gemini, Kimi and Claude models
According to daily.dev’s summary of the October 2 changes, the affected models and suggested replacements are:
| Deprecated model | Suggested replacement |
|---|---|
| Gemini 3.5 Flash and Gemini 3.6 Flash | Gemini 3.8 Flash |
| Kimi K2.7 Code | Kimi K3 |
| Claude Opus 4.7 | Claude Opus 5 |
The scope extends beyond the Chat model selector. GitHub’s notice explicitly includes inline edits, ask and agent modes, and code completions. A review limited to developers’ Chat selections would therefore cover only part of the announced change.
October 19: another six models scheduled
daily.dev’s mid-October report lists Gemini 3.7 Flash, GPT-5.5, GPT-5.4, GPT-5.4 mini, GPT-5 mini and Grok 4.5 for deprecation on October 19, 2026, across the same Copilot experiences.
The report names Gemini 3.8 Flash, GPT-5.6 Sol, GPT-5.6 Luna and Grok 4.6 as suggested alternatives. It specifically identifies GPT-5.6 Luna as the replacement for both GPT-5.4 mini and GPT-5 mini.
These are migration recommendations. The supplied evidence contains no comparative benchmarks, pricing or performance measurements for the replacement models, so it does not establish equivalent behavior or a quantified improvement over the retiring options.
Check model policies before assuming access
The two daily.dev summaries give different levels of detail about administrative action. The October 2 coverage says Copilot Enterprise and Business administrators may need to enable alternatives through Copilot model policies. It advises checking that the alternative model policy is enabled in organization or enterprise Copilot settings; once enabled, the new model appears in the Copilot Chat model selector.
The October 19 coverage says suggested alternatives are automatically enabled for Enterprise and Business customers under default model enablement, unless an administrator has disabled them. For GPT-5.6 Luna specifically, it says automatic enablement depends on neither the global default nor the model itself having been disabled.
The distinction matters: the earlier summary emphasizes a possible manual step, while the later one describes conditional automatic enablement. Neither supports assuming that every organization already has access to every suggested replacement. The practical implication is to verify the applicable settings and model availability rather than treating the replacement list alone as proof of access.
Automatic removal still leaves migration work
Both daily.dev reports say administrators do not need to remove deprecated models after the cutoff. They also advise updating workflows and integrations to use supported models before the deprecation date. The mid-October report says deprecated models will no longer be available in the Copilot Chat model selector or other Copilot experiences.
For teams reviewing the changes after the announced dates, those instructions translate into three checks:
- Identify workflows and integrations that still refer to the listed models.
- Update those choices to supported alternatives, using the reported replacement guidance where applicable.
- Verify the relevant organization or enterprise model policies and confirm that users can see the enabled alternative in Copilot Chat.
Automatic removal addresses the retired model’s availability. Updating workflows and checking access to a replacement remain separate administrative tasks. The evidence supports that migration checklist, but leaves replacement performance and confirmation of the October 19 rollout unresolved.
