Model release · Microsoft
MAI-Code-1.1-Flash
- Released
- Review updated
Vs 1-Flash: Terminal-Bench 2.1 +22%, about +25% token efficiency, price about 1/4. The October 7 update makes a 3-bit quantized version available for local use with the full 256K context window and no local inference charges. Microsoft recommends more than 120GB of RAM. Experimental local access in the Copilot app, CLI and VS Code is planned by the end of October, not already delivered.
Availability
GitHub Copilot; downloadable local inference