GitHub announced in its Changelog on October 2, 2026 that four models are deprecated across all GitHub Copilot experiences — Copilot Chat, inline edits, ask and agent modes, and code completions. The retirees: Gemini 3.5 Flash, Gemini 3.6 Flash, Kimi K2.7 Code, and Claude Opus 4.7. GitHub names one suggested alternative for each: Gemini 3.8 Flash for both Flash retirees, Kimi K3 for K2.7 Code, and Claude Opus 5.5 for Opus 4.7.

This is wave one of a two-part cleanup. Wave two lands October 19, and GitHub published that list back on September 18 — six more models, with GPT-5.4 and GPT-5.5 among them. Copilot is now officially on a quarterly-ish pruning schedule, and pinned model names have become a liability.

The retirement list#

Retired today (October 2): Gemini 3.5 Flash → Gemini 3.8 Flash; Gemini 3.6 Flash → Gemini 3.8 Flash; Kimi K2.7 Code → Kimi K3; Claude Opus 4.7 → Claude Opus 5.5. Retiring October 19 (per GitHub's September 18 notice): 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.

Read GitHub's wording literally: "no action is required to remove the models after they have been deprecated." That is about cleanup, not rerouting — neither notice says Copilot will quietly map a saved selection onto the suggested replacement. Anything on your side that names the old model by string is yours to fix.

A developer workspace with glowing arrows swapping old sticky notes for bright new cards — migrating Copilot models
Swapping models is not just a name change: each replacement sits on GitHub's rate card at its own price. AI-generated illustration.

What the swap does to your bill#

GitHub names a replacement per model and says nothing about price. An independent analysis published October 2 lined every swap up against GitHub's own per-token rate card (fetched the same day), and two swaps get more expensive. The painful one: Kimi K2.7 Code ran at $0.95 input / $4.00 output per million tokens; Kimi K3 lists at $3.00 / $15.00 — input more than triples. GPT-5.4 users sent to GPT-5.6 Sol also pay more: $2.50 / $15.00 climbs to $4.00 / $20.00.

Not everything moves up. Gemini 3.8 Flash's $0.75 / $3.75 rate is promotional through December 31, 2026 — GitHub does not say what happens after. Claude Opus 4.7 users actually get a cheaper upgrade than GitHub's own suggestion: the same analysis found Opus 5.5 at $4.00 / $20.00, undercutting the suggested Opus 5's $5.00 / $25.00. And tab-complete is untouched by all of this: code completions and next-edit suggestions stay unlimited on every paid plan.

A glowing terminal window with an AI assistant figure and pricing tags drifting upward
The retirements only move the chat, agent, and CLI side of the bill — completions stay unlimited. AI-generated illustration.

What to check today#

Old model names hide in four places: the COPILOT_MODEL environment variable (CI scripts are the usual offender), a model key in ~/.copilot/settings.json or the committed .github/copilot/settings.json, custom agent profiles' model frontmatter, and per-subagent modelPolicy overrides. Sweep all of them now — including the wave-two models, before October 19 forces the issue.

For Business and Enterprise admins: the wave-one notice only says admins "may need to enable access to the alternative models through their model policies." If your org turned off the global default, the replacements do not appear on their own — open Copilot model settings today and confirm Gemini 3.8 Flash, Kimi K3, and Claude Opus 5 are enabled. The alternative for teams tired of chasing deprecation notices: point shared configs at auto and stop pinning.