As early as June this year, Microsoft first announced MAI-Code-1. This coding model is optimized for GitHub Copilot workloads, with an emphasis on reasoning efficiency. Later, the Flash version of MAI-Code-1 arrived in GitHub Copilot as one of the models available to users, competing with models from OpenAI and Anthropic.
When MAI-Code-1-Flash was first released, its pricing and efficiency were highly competitive. However, it was soon overshadowed by lower-priced and high-performing Chinese models such as GLM-5.2 and Kimi K3, as well as OpenAI’s GPT-5.6 Luna.
Today, Microsoft announced the upgraded MAI-Code-1.1-Flash. In addition to improving coding capabilities, it significantly reduces costs and Token consumption. Microsoft said that when MAI-Code-1.1-Flash is used through GitHub Copilot CLI, its performance in the Terminal-Bench 2.1 test improves by 22%, while performance on .NET-related tasks improves by 15%.
Microsoft also said that the new model generates Tokens 25% faster, while the number of Tokens required to complete the same tasks has decreased by 25%. Thanks to improvements in training and serving efficiency, Microsoft has reduced the price of MAI-Code-1.1-Flash to one-quarter that of the initial model. For annual Copilot subscribers, the model has a premium request multiplier of 0.25x.
According to GitHub Copilot’s pricing table, MAI-Code-1.1-Flash costs $0.20 per 1 million input Tokens (approximately 1.4 yuan at the current exchange rate), $0.02 per 1 million cached input Tokens (approximately 0.14 yuan at the current exchange rate), and $1.20 per 1 million output Tokens (approximately 8.1 yuan at the current exchange rate).
By comparison, MAI-Code-1-Flash costs $0.75 (approximately 5.1 yuan at the current exchange rate), $0.075 (approximately 0.51 yuan at the current exchange rate), and $4.50 (approximately 30.4 yuan at the current exchange rate) per 1 million Tokens, respectively. The new model’s price is therefore significantly lower.
In addition to improved coding capabilities, MAI-Code-1.1-Flash adds native vision capabilities, allowing it to understand and analyze image content.
MAI-Code-1.1-Flash is now being rolled out across GitHub Copilot services, including VS Code, Visual Studio, JetBrains IDEs, Copilot CLI, GitHub Mobile, and other supported platforms.
Free and student users can access the model through automatic model selection, while paid users can manually select MAI-Code-1.1-Flash.
With the upgraded model officially launched, GitHub will stop offering MAI-Code-1-Flash across all Copilot services on September 10, 2026.