GitHub Copilot retires six models today, October 19, including GPT-5.5, GPT-5.4 and Grok 4.5. GitHub’s changelog says the models are deprecated across all Copilot experiences — Copilot Chat, inline edits, ask and agent modes, and code completions — with a named replacement for each one.
The retirement was announced on September 18, giving teams a month to move. From today, workflows and integrations that still point at the six models need to be on the supported alternatives GitHub lists.
The Six Models Leaving Copilot Today
The departing models are Gemini 3.7 Flash, GPT-5.5, GPT-5.4, GPT-5.4 mini, GPT-5 mini and Grok 4.5. GitHub’s notice is explicit that this is a Copilot deprecation: the models disappear from Copilot’s surfaces and model selector, and no action is required from users to remove them after the date.
For developers who had settled on one of these models for daily work — its behaviour in agent mode, or its handling of long editing sessions — the change is not invisible. The replacement is a different model, with its own strengths, pricing and failure modes, and GitHub’s table is a migration instruction rather than a promise of identical behaviour.
That makes today a testing deadline as much as a switch-off date. A model selector can show the suggested replacement while a team’s pinned workflow still names the old model in a configuration file, an evaluation harness or an internal runbook. The picker changing does not audit those places. Teams that rerun a real task — their own code, their own review standards — against the replacement before trusting it will find the differences today, on their own terms, rather than in production later this week.
What Replaces Each Model
GitHub names one suggested alternative per model. GPT-5.5 and GPT-5.4 both move to GPT-5.6 Sol. GPT-5.4 mini and GPT-5 mini move to GPT-5.6 Luna. Gemini 3.7 Flash is replaced by Gemini 3.8 Flash, and Grok 4.5 by Grok 4.6.
Once enabled, users will see the replacement in the Copilot Chat model selector in supported experiences. GitHub’s advice is direct: “Please update your workflows and integrations to use the supported models before this date” — which, as of today, means now.
What Business And Enterprise Admins Need To Check
Under default model enablement, the suggested alternatives are automatically enabled for Copilot Enterprise and Copilot Business customers — unless an administrator has turned off the global default or explicitly disabled the replacement model. Admins who have done either need to enable access through the model policies in Copilot settings, or their users will find the old models gone and the new ones missing.
GitHub Enterprise customers with questions or concerns are encouraged to contact their account team. The practical audit is the same one that applies to next month’s Copilot retirement wave: search configurations, wrappers, evaluations and runbooks for the six retiring names, and test the replacements against real development work rather than a generic prompt.
Why It Matters
GitHub Copilot is where most developers meet model change as a deadline rather than a headline: a model picker is a dependency, and today six of those dependencies expire at once. The pattern — one month’s notice, a named successor, automatic enablement unless an admin opted out — is becoming the standard lifecycle for the AI pair programmer with 50 million users, and it shifts the cost of model progress onto every team that pinned its tooling to a specific model name inside Microsoft Copilot’s wider product family.
Background
GitHub has been rotating the third-party models inside Copilot — from OpenAI, Anthropic, Google and xAI — in waves through 2026, as each lab ships successors and older versions are wound down. This October 19 batch is the third wave in under two months, and another follows on November 16, when Claude Opus 4.8, Opus 5 and GPT-5.6 Terra leave the picker. The direction is steady: Copilot’s model menu is a moving shelf, and the changelog is where its expiry dates live.
0 Comments