Share with your CTO
GitHub is deprecating six models across all Copilot experiences on October 19th, 2026, including GPT-5.5, GPT-5.4, GPT-5.4 mini, GPT-5 mini, Gemini 3.7 Flash, and Grok 4.5. The full deprecation notice maps each model to a suggested replacement, ranging from GPT-5.6 Sol and GPT-5.6 Luna to Gemini 3.8 Flash and Grok 4.6. Enterprise and Business customers on default settings get the replacements automatically, but admins who’ve disabled global defaults need to manually enable the alternatives through Copilot settings before the deadline.
What this means for your business
Any engineering team that has hardcoded a specific model name into CI/CD pipelines, IDE configurations, or API integrations faces a hard stop on October 19th. The “default model enablement” safety net only catches teams whose admins never locked down model access. If your organization applied any governance controls to restrict which models developers could use, those same controls now block the automatic upgrade path. Someone needs to audit that before the deadline, not after.
The churn rate of model versions in GitHub Copilot is accelerating. GPT-5.4 and GPT-5.5 are being retired within what is almost certainly months of their introduction, replaced by GPT-5.6 variants named “Sol” and “Luna.” That naming convention signals GitHub is moving toward branded model tiers rather than simple version numbers, which matters for how enterprises write model policies. A policy pinned to a version number becomes stale on a quarterly cadence. A policy built around capability tiers or provider families holds longer. Teams still writing “GPT-5.4” in their configs are building technical debt faster than they realize.
The signal worth watching: how GitHub handles model governance at this pace will determine whether enterprise IT can maintain meaningful oversight of AI tooling at all. When six models deprecate simultaneously across chat, inline edits, code completions, and agent modes, the blast radius of a missed update is no longer a developer annoyance. It’s a workflow outage. The tradeoff is real: faster model improvement cycles benefit developers who want the best available capability, but they compress the window for IT to validate, approve, and document each new model before it lands in production use.
Based on reporting from Upcoming deprecation of selected GitHub Copilot models in mid-October, originally published 2026-09-18 15:00:00.
