Microsoft Slashes AI Coding Costs With MAI-Code-1.1-Flash

Microsoft has launched MAI-Code-1.1-Flash, a new coding model for GitHub Copilot that promises stronger performance, faster responses, and significantly lower costs.

The model builds on MAI-Code-1, which Microsoft introduced in June, and the later MAI-Code-1-Flash release.

Microsoft says the latest version improves coding results while using fewer tokens for the same workloads.

MAI-Code-1.1-Flash improves coding performance

MAI-Code-1.1-Flash scores 22% higher on Terminal-Bench 2.1 when developers use it through the GitHub Copilot CLI. Microsoft also reports a 15% improvement on .NET coding tasks.

Token streaming runs 25% faster than with the previous model, while the model uses around 25% fewer tokens to complete comparable tasks.

Microsoft has also added native vision support, allowing MAI-Code-1.1-Flash to analyze and understand images as part of coding workflows.

Microsoft cuts MAI-Code pricing

Microsoft says improvements to training and serving have reduced the cost of MAI-Code-1.1-Flash to roughly one-quarter of the original model.

Input tokens cost $0.20 per 1 million tokens, cached input costs $0.02, and output costs $1.20 per 1 million tokens.

MAI-Code-1-Flash costs $0.75 for input, $0.075 for cached input, and $4.50 for output per 1 million tokens.

Annual GitHub Copilot subscribers also get a 0.25x premium request multiplier when using the new model.

GitHub Copilot rollout is underway

MAI-Code-1.1-Flash is rolling out across VS Code, Visual Studio, JetBrains IDEs, GitHub Copilot CLI, and GitHub Mobile.

Free and Student users can access it through automatic model selection, while paid subscribers can manually select the model.

GitHub plans to remove the older MAI-Code-1-Flash model from all Copilot experiences on September 10, 2026.

Microsoft is also expanding its Flash model strategy elsewhere. MAI-Cyber-1-Flash is expected to cut MDASH costs by 50%, and the company has recently introduced the MAI-Image-2.6 model.