Tasks and inputs
MAI Cyber 2 Flash is listed for Text generation, Coding & Code Generation. It accepts text and produces text. Use the publisher’s input format and preprocessing for your chosen deployment.
SyncDev model directory · Profile updated Nov 18, 2026
MAI Cyber 2 Flash is Microsoft AI's security model for vulnerability discovery and threat investigation, with a 1M-token context, shipping in MDASH.
The specification score summarizes recorded features and limits. It is not a benchmark result.
Pricing conditions: Indicative list price: $0.10 input / $0.01 cached / $0.50 output per 1M tokens. Billed through Microsoft security products; confirm with your licensing.
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
You will not find MAI Cyber 2 Flash in a model picker. Microsoft AI announced it at Ignite on 17 November 2026 as a specialist for defenders, and it ships only inside Microsoft's security products, starting with Codename MDASH, the company's agentic code-security system. Eligible customers get it automatically from mid-December 2026, when it replaces the current model in the MAI profile.
That makes this page less a buying guide than a briefing on what is about to change under you.
Long codebases are the reason for the biggest spec change. The context window grows to 1,000,000 tokens, four times that of MAI Cyber 1.1 Flash, which shipped only on 27 October.
| Token type | Per 1M |
|---|---|
| Input | $0.10 |
| Cached input | $0.01 |
| Output | $0.50 |
Microsoft says output is 86% cheaper than 1.1 Flash. Because you consume the model through a security product rather than an API, how that list price reaches your invoice depends on your licensing. Treat it as indicative and confirm with your account team.
On CyberGym-E2E, Microsoft reports 92.3% task success against 91.5% for GPT-6 Astra, using half the reasoning budget and under 2% of the cost per task. It also excluded 92 of 905 tasks that overlapped with training data, leaving 813, which is a disclosure worth crediting. On WebExploitBench it reports 40.4% against 38.4% for GPT-5.6 Cyber.
Both figures are publisher-reported. There is also a loose end. The comparison in the body text is with GPT-5.6 Cyber, while a footnote talks about GPT-6 Cyber and promises results later. Until those appear, there is no published head-to-head with OpenAI's current cyber model.
If MDASH already runs on your code, capture a baseline now. Export last month's findings, note which ones your team confirmed and which were noise, and repeat the same scans once 2 Flash takes over. Precision on your own repositories is the only number that matters here. For broader context, see the Microsoft model lineup.
Compare the model on the work you need to complete.
MAI Cyber 2 Flash is listed for Text generation, Coding & Code Generation. It accepts text and produces text. Use the publisher’s input format and preprocessing for your chosen deployment.
Documented limits: 1M context. The listed route has starting rates of $0.10 input and $0.50 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 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).
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.
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.