MAI-Code-1.1-Flash review, pricing and limits

On benchmarks, MAI-Code-1.1-Flash scores 72.6% on SWE-bench Verified, averaging 8.6K tokens per solved task, and 62.9% on Terminal-Bench 2.1 inside GitHub Copilot CLI, averaging 17.0K tokens, per Microsoft's model card. In that card's own runs the June MAI-Code-1-Flash managed 71.6% and 51.7%, Claude Haiku 4.5 69.8% and 49.4%, and GPT-5.4 mini 69.2% and 60.7%, so the terminal result is the real step forward and the SWE-bench lead over those peers is a point or three. Microsoft also reports a 15% gain on .NET-specific tasks and roughly 25% fewer tokens per completed task than the June release. For image-to-code work, the card records 74.1% on Text2WebApp and 42.1% on ScreenShot2WebApp, both internal Microsoft tests.

Context window sits at 256,000 tokens with a 128,000-token maximum output ceiling, unchanged from the predecessor. Microsoft reports 25% faster token streaming inside GitHub Copilot's CLI, which combined with the lower token count per solve shortens the wall-clock time of longer agentic sessions. Teams comparing agentic coding tools more broadly, including terminal-native options, can see how it stacks up in Hokai's Cursor Composer vs. Claude Code comparison and its locally runnable coding models roundup.

Running it on your own machine became an option on October 7, 2026. Microsoft's Windows Command Line team describes a quantized build at roughly 3.3 bits per weight that takes 53GB, 80% smaller than the cloud version, and peaked at 75.5GB of memory at the full 256K context on a Surface Laptop Ultra with NVIDIA RTX Spark. In Microsoft's October 5 tests that local build scored 70.8% on SWE-bench Verified and 66.3% on Terminal-Bench 2.1, against 32.0% and 23.6% for OpenAI's larger gpt-oss open-weight model on the same harness. Microsoft charges nothing per local call and says experimental access reaches the GitHub Copilot app, Copilot CLI and VS Code by the end of October 2026, through Copilot's Auto routing or by choosing it under the Windows ML provider. Developers used to pulling open-weight coders into Ollama or LM Studio should note the difference: Microsoft has published no Hugging Face repository or open-weight license for this model, so GitHub Copilot is the supported local path.

Inputs are text and, new in this release, native images: MAI-Code-1.1-Flash can read screenshots, diagrams, and UI mockups directly, a capability the text-only June release lacked. Output stays text and code, with native function calling for tool use. There is no toggleable extended-reasoning mode; Microsoft routes deliberate multi-step planning to the separate MAI-Thinking-1 model, keeping this one tuned for single-pass code generation, repository question answering, and refactoring.

Pricing dropped sharply from the June release: Microsoft cut the rate across every tier of the per-token structure, framing the update as delivering more capability for meaningfully less spend. The model bills through GitHub Copilot and through Microsoft Foundry directly. Exact current rates and worked cost examples live in the pricing FAQ below.

MAI-Code-1.1-Flash is live across VS Code, Visual Studio, JetBrains IDEs, Eclipse, Xcode, GitHub Mobile, Copilot CLI, the Copilot cloud agent, and Copilot Chat on GitHub, spanning every Copilot plan from Free through Enterprise, see what Microsoft Copilot actually gets you across its different products. Free and Student users get it only through automatic model selection; paid tiers can select it manually. Business and Enterprise administrators have to switch on its Copilot policy themselves, since it ships off by default at the organization level. Outside Copilot it is reachable through Microsoft Foundry on Azure; OpenRouter's catalog listed no MAI model when Hokai checked in mid-October 2026, and no AWS Bedrock or Google Vertex listing has surfaced.

On safety, Microsoft's published model card describes filtering harmful content out of the pre-training mixture, then applying supervised fine-tuning and reinforcement learning to shape model behavior toward safe, helpful output. Weights are not open, local build included, and Microsoft discloses a pretraining data cutoff of December 2025 for this 1.1 refresh.

The model fits teams already running on GitHub Copilot who want an inexpensive, fast agent for CLI-heavy or .NET-heavy work, and for high-volume tasks like bulk refactors or repo Q&A where a frontier-priced model is overkill; Hokai's best AI coding assistants roundup and free coding-question tools guide cover the wider field for teams still choosing. Teams that need deliberate, multi-step reasoning are better served by MAI-Thinking-1 or a frontier coding model, and DeepSeek-V4-Flash-0731, from DeepSeek, posts a stronger terminal-benchmark score at a lower blended price for teams not tied to the Copilot ecosystem, while Kimi K2.7-Code, its lower-latency Kimi K2.7 Code HighSpeed variant, and Google's Gemini 3.5 Flash are the closest open-weight, speed-tier, and budget-tier alternatives respectively; GPT-5.6 Luna is the closest OpenAI-ecosystem pick, and Hokai's guide to choosing an AI assistant covers the wider field, browsable in full in the models directory. MAI-Code-1-Flash, the June predecessor, was retired across all GitHub Copilot surfaces on September 10, 2026, and Microsoft has described 1.1 as a drop-in replacement at the prompt level rather than a rewrite.