Gemini 2.5 Pro and Gemini 3 Flash deprecated
By Dillip Chowdary • Aug 03, 2026 • Source: GitHub Changelog
As of July 31, 2026, GitHub has deprecated Gemini 2.5 Pro and Gemini 3 Flash across all GitHub Copilot experiences. The change covers Copilot Chat, inline edits, ask and agent modes, and code completions. The notice was published on the GitHub Blog via the GitHub Changelog.
Deprecation here is product-wide rather than limited to a single surface. The same two models are removed from chat, edit, ask, agent, and completion paths, so any workflow that pinned or defaulted to Gemini 2.5 Pro or Gemini 3 Flash inside Copilot is affected at once.
Advertisement
Tech Pulse Daily
Get tomorrow's pulse first
Join engineers who read Tech Pulse before stand-up. Free, weekday mornings.
For engineers and builders, that means existing Copilot habits, prompts, and agent setups that assumed those models are available need a model switch. Teams that standardized on one of them for chat, inline edits, or agent runs should confirm which model Copilot now routes to and retest those flows.
In market terms, GitHub is pruning two named Gemini models from the Copilot model set on a fixed date rather than leaving them as optional backends. That tightens the choice set for users who preferred those specific Gemini options inside the Copilot product family.
What to do next: check your Copilot model selection for Chat, inline edits, ask, agent, and completions after July 31, 2026, replace any remaining use of Gemini 2.5 Pro or Gemini 3 Flash, and re-run critical edit and agent tasks against the models GitHub still offers. Watch follow-up changelog entries for replacement defaults and any further Gemini deprecations.
Advertisement
🔎 More interesting news
- AI price wars: OpenAI cuts GPT-5.6 Luna prices by 80% as model competition shifts toward…
- Agentic coding goes hands-free as OpenAI brings GPT-Live's full duplex voice control to…
- AI Gateway: GPT-5.6 pricing and speed updates
- Thinking Machines debuts Inkling Small open source AI model nearing performance of…
- Today's full Tech Pulse briefing →