Model Availability
Six Copilot Models Have an October 19 Deadline
October 19 is the scheduled change date, not September 18. GitHub's notice lists six models for removal across Copilot experiences and suggests replacements. Teams have a migration task now; the notice does not say those models have already disappeared.
Citation-ready: GitHub schedules six Copilot model deprecations for October 19, 2026; a suggested replacement still needs to be enabled and checked in the intended workflow.
Evidence boundary: GitHub-specific scheduled availability change, not a claim about retirement in other providers' APIs or an independent comparison of replacement quality. No organization setting changed.
What happened and why it matters
A supported model appearing in a catalog does not establish that an organization permits it or that an existing workflow has stopped requesting the old one.
Primary evidence
Primary reference: GitHub September 18 model deprecation notice. Kaleido Field checked the event date and the article's attributed facts against this source.
| Source date | September 18, 2026 notice; October 19, 2026 scheduled deprecation |
|---|---|
| Checked by Kaleido Field | September 20, 2026, CST |
| Source function | model availability -> migration timing and access policy |
The replacement map is explicit
GitHub suggests Gemini 3.8 Flash for Gemini 3.7 Flash; GPT-5.6 Sol for GPT-5.5 and GPT-5.4; GPT-5.6 Luna for GPT-5.4 mini and GPT-5 mini; and Grok 4.6 for Grok 4.5. All six listed dates are October 19, 2026.
The notice says suggested alternatives follow default enablement for Business and Enterprise unless administrators disabled the default or the particular model. Check the policy that actually applies to the user rather than assuming the recommendation grants access.
Validate the workflow that will run after the deadline
Inventory explicit model references in integrations and saved configurations, then exercise a representative permitted task with the replacement. Keep output correctness, latency and any changed behavior separate from the fact that the request completed.
This is a maintenance check, not a model ranking. A general recommendation cannot establish suitability for every repository or instruction set. The parallel-work report illustrates why the integrated result still matters after individual agent tasks succeed.
Evidence boundary
GitHub-specific scheduled availability change, not a claim about retirement in other providers' APIs or an independent comparison of replacement quality. No organization setting changed.
FAQ
Does this notice retire the same models from every external API?
No. Its stated scope is GitHub Copilot experiences.