Tasks and inputs
MAI Code 2 is listed for Image-text-to-text, Text generation, Coding & Code Generation. It accepts text, image and produces text. Use the publisher’s input format and preprocessing for your chosen deployment.
SyncDev model directory · Profile updated Nov 18, 2026
MAI Code 2 is Microsoft AI's coding model for multi-file work, debugging and frontend builds, live in GitHub Copilot at $0.50/$2.50 per 1M (promo).
The specification score summarizes recorded features and limits. It is not a benchmark result.
Pricing conditions: Indicative GitHub Copilot usage-based rates: $0.50 input / $0.025 cached / $2.50 output per 1M until 17 May 2027; then $1.00 / $0.05 / $5.00. Confirm on your Copilot billing page.
Recorded modality support shows what this model can accept and produce. An undocumented modality is shown as unknown rather than assumed to be unsupported.
Accepts
Produces
Microsoft AI's pitch for MAI Code 2 fits in one comparison: most of a frontier coder's output for a fraction of the bill. It launched at Ignite on 17 November 2026 inside GitHub Copilot, and it is aimed squarely at the unglamorous middle of enterprise engineering, the debugging, refactors and multi-file changes that fill a normal sprint.
The predecessor line was Flash-only. MAI-Code-1.1-Flash was a light model; this is the first full-size step.
Copilot bills it per token under usage-based billing. The current rates are promotional and expire on 17 May 2027.
| Until 17 May 2027 | From 17 May 2027 | |
|---|---|---|
| Input (per 1M) | $0.50 | $1.00 |
| Cached input | $0.025 | $0.05 |
| Output (per 1M) | $2.50 | $5.00 |
Plan around the second column. Microsoft's claim that the model costs about 40% of GPT-6.1 Sol is calculated at the 50% promotional price, so at list rates the ratio roughly doubles. These are indicative figures; your Copilot billing page is the authority.
Microsoft reports that, averaged across DeepSWE 1.1, Terminal Bench 2.1, SWE-Atlas and FrontierCode v1.1, MAI Code 2 reaches 98% of GPT-6.1 Sol and 96% of Claude Opus 5.5. In a blind frontend comparison with 258 annotators, reviewers preferred its output over GPT-6 Sol 64% of the time. All of that is publisher-reported, measured at maximum reasoning effort through the Copilot CLI without subagents.
Microsoft is also unusually direct about the limits. It says the model is not yet tuned for fundamental science research or very long-horizon agentic work. Believe that part.
One gap matters. Neither Microsoft nor GitHub has published a context window, and third-party trackers disagree wildly. Do not size a repo-wide prompt around a number you have not confirmed in your own session.
The Flash sibling is roughly a tenth of the price and is meant for explanations and quick edits. Give a real refactor ticket to both. If Flash's diff needs more than one round of fixing, the bigger model is the cheaper choice. Compare against other options on the coding models page.
Compare the model on the work you need to complete.
MAI Code 2 is listed for Image-text-to-text, Text generation, Coding & Code Generation. It accepts text, image and produces text. Use the publisher’s input format and preprocessing for your chosen deployment.
The listed route has starting rates of $0.50 input and $2.50 output per million tokens. Measure a completed task using the prompts, documents and tools your application will use. Include reasoning and retries in the result.
Use the model identifier and account requirements for the selected API route. Confirm its request limits and output format with one complete task before connecting a larger workflow.
These records share a directory category. Check the source card, output type and deployment requirements before treating a model as a replacement.
Microsoft
MAI Code 2 Flash is Microsoft AI's low-cost coding model for edits, fixes and explanations, on every GitHub Copilot plan including Free and Student.
Microsoft
MAI Cyber 2 Flash is Microsoft AI's security model for vulnerability discovery and threat investigation, with a 1M-token context, shipping in MDASH.
Gemini 4 Hydrogen is Google's fast, lower-cost Gemini 4 model for everyday coding and agentic execution, with a 1M context, xhigh thinking and a free tier.
OpenAI
GPT-6 Cyber is OpenAI's restricted security model for authorised vulnerability investigation, reproduction and fix verification, for Daybreak Red customers.
ricursive
Ripple-65B Preview is Ricursive Intelligence's first open model: a dense 65B coding-focused transformer trained from scratch, with a 1M-token context.
particle-ai
Particle-1 is Particle's first software-engineering model, an RL-tuned DeepSeek V4.1 Flash that aims to solve coding tasks with fewer tokens.