Copilot's MAI-Code-1.1-Flash goes GA: 73% cheaper, adds vision
Microsoft's small coder hits every Copilot editor at a 0.25x multiplier; Ollama local models and cross-session memory also land in Copilot for JetBrains.
Copy markdown
73% cheaper - and it can finally see
MAI-Code-1.1-Flash is generally available across Copilot CLI, VS Code, Visual Studio, JetBrains, Xcode, Eclipse, GitHub Mobile and github.com. Microsoft's small coder costs 73% less than the model it replaces, adds native image understanding, and bills at just a 0.25x premium multiplier for usage-based subscribers.
Mark the calendar: 1-Flash retires Sept 10
GitHub will deprecate the older MAI-Code-1-Flash across all Copilot experiences on September 10, 2026. Repoint any pinned model configs to 1.1-Flash now; Enterprise and Business admins may first need to enable it via model policies.
Run local models inside Copilot for JetBrains
The latest JetBrains plugin adds Ollama as a bring-your-own-key provider, so Copilot Chat can point at models running on your own machine. That is a real unlock for offline, air-gapped, or cost-sensitive work you could not route through the cloud before.
Copilot Memory lands in JetBrains too
A new toggle in the Copilot settings portal lets Copilot retain project details and preferences across chat sessions, so you stop re-explaining your stack every conversation.
See exactly what is burning your AI credits
The downloadable usage report now itemizes each model by input, output, cache-read and cache-write tokens next to the credits they consumed - on Business, Enterprise and individual plans. A surprise bill is finally explainable, and shrinkable.