# MAI Cyber 2 Flash

> Microsoft AI's specialist security model for vulnerability discovery and investigation, with a 1M-token context window, announced November 17, 2026 for Microsoft security products.

- Maker: Microsoft
- Line: MAI Cyber (security models)
- Canonical page: https://ai-tldr.dev/models/mai-cyber-2-flash/
- Updated: 2026-11-17

## In short

MAI Cyber 2 Flash is a specialist security model that Microsoft AI announced at Ignite on November 17, 2026. It has a 1M-token context window and costs $0.10 per million input tokens and $0.50 per million output tokens.

## Facts

| Fact | Value | Source |
| --- | --- | --- |
| Context window | 1M tokens | [1] |
| Input price | $0.10 / 1M tokens (cached $0.01) | [2] |
| Output price | $0.50 / 1M tokens | [2] |
| Licence | Proprietary · API only | [1] |
| Released | 17 November 2026 | [3] |

## Sources

1. [MAI Cyber 2 Flash model page](https://microsoft.ai/models/mai-cyber-2-flash/)
2. [MAI Cyber 2 Flash: Leading security. Lower cost (launch post)](https://microsoft.ai/news/introducing-mai-cyber-2-flash/)
3. [MAI Cyber 1.1 Flash: Sharper findings and fewer tokens inside MDASH](https://microsoft.ai/news/mai-cyber-1-1-flash/)

## Overview

MAI Cyber 2 Flash is a specialist security model that Microsoft AI announced at Ignite on November 17, 2026. Microsoft built it on its own security data and expertise to find vulnerabilities, recommend fixes and investigate threats around the clock — the monitoring and patching work of security agents — at a fraction of the cost of larger general models. It follows MAI Cyber 1 Flash (August 2026) and MAI Cyber 1.1 Flash (October 2026), both of which ran inside Microsoft's MDASH security agent.

Its context window is one million tokens, four times that of MAI Cyber 1.1 Flash, so it can hold more of a codebase during long, multi-step investigations. Microsoft lists its price at $0.10 per million input tokens, $0.01 per million cached input tokens and $0.50 per million output tokens.

On CyberGym-E2E, a benchmark of real-world vulnerability discovery, Microsoft reports 92.3% task success against 91.5% for GPT-6 Astra, using half the reasoning budget and at less than 2% of GPT-6 Astra's cost per task. On WebExploitBench it scored 40.4% against 38.4% for GPT-5.6 Cyber at the same reasoning budget. Microsoft found that 92 of CyberGym-E2E's 905 tasks drew on the same open-source projects and vulnerability reports as some of its training data, so it excluded them for every model and reports results on the remaining 813. It has not yet compared the model with GPT-6 Cyber, which it says is available only to select customers.

MAI Cyber 2 Flash is not offered as a general API. Microsoft is bringing it exclusively to its own security products, including Codename MDASH, where it replaces the existing model in the MAI profile for eligible customers starting in mid-December 2026.

## Benchmarks

- CyberGym-E2E (813 tasks): 92.3% ([source](https://microsoft.ai/news/introducing-mai-cyber-2-flash/))
- WebExploitBench: 40.4% ([source](https://microsoft.ai/news/introducing-mai-cyber-2-flash/))

## Pricing

- Input: $0.10 / 1M tokens
- Cached input: $0.01 / 1M tokens
- Output: $0.50 / 1M tokens

List price Microsoft used for its cost-per-task comparison; the model ships inside Microsoft security products.

Source: https://microsoft.ai/news/introducing-mai-cyber-2-flash/

## Strengths

- 92.3% task success on CyberGym-E2E (813 tasks) against 91.5% for GPT-6 Astra, at half the reasoning budget, per Microsoft
- Less than 2% of GPT-6 Astra's cost per task on CyberGym-E2E at list prices
- 40.4% on WebExploitBench against 38.4% for GPT-5.6 Cyber at the same reasoning budget
- One-million-token context window, four times that of MAI Cyber 1.1 Flash
- Microsoft reports performance comparable to leading models on investigation-agent and detection-authoring tasks in early Microsoft Perception evaluations

## Best for

- Reach for it inside Microsoft's security products, such as Codename MDASH, for continuous vulnerability discovery and patch recommendations across a codebase.
- Reach for it for long, multi-step security investigations that need a large share of a codebase in context.
- Look elsewhere if you need a security model through a general-purpose API — Microsoft offers MAI Cyber 2 Flash only inside its own security products.

## MAI Cyber (security models) versions

- MAI Cyber 2 Flash (17 November 2026): 1M-token context, 92.3% on CyberGym-E2E; coming to Microsoft security products including MDASH. — this page
- MAI Cyber 1.1 Flash (27 October 2026): Sharper findings and fewer tokens inside MDASH.
- MAI Cyber 1 Flash (13 August 2026): First MAI Cyber model, introduced inside MDASH.

## FAQ

### What is MAI Cyber 2 Flash?

It is Microsoft AI's specialist security model, announced on November 17, 2026, for finding vulnerabilities, recommending fixes and investigating threats. It has a one-million-token context window.

### How does MAI Cyber 2 Flash compare with GPT-6 Astra?

On CyberGym-E2E, Microsoft reports 92.3% task success for MAI Cyber 2 Flash against 91.5% for GPT-6 Astra, using half the reasoning budget and at less than 2% of the cost per task. These are Microsoft's own results on 813 of the benchmark's 905 tasks.

### Can I call MAI Cyber 2 Flash through an API?

Not as a general API. Microsoft says it is coming exclusively to Microsoft security products, including Codename MDASH, where it replaces the existing MAI-profile model for eligible customers from mid-December 2026.

### How much does MAI Cyber 2 Flash cost?

Microsoft's list price is $0.10 per million input tokens, $0.01 per million cached input tokens and $0.50 per million output tokens.

## Links

- [MAI Cyber 2 Flash: Leading security. Lower cost (launch post)](https://microsoft.ai/news/introducing-mai-cyber-2-flash/)
- [MAI Cyber 2 Flash model page](https://microsoft.ai/models/mai-cyber-2-flash/)
- [MAI Cyber 1.1 Flash: Sharper findings and fewer tokens inside MDASH](https://microsoft.ai/news/mai-cyber-1-1-flash/)
- [Seven new world class models at scalable price-performance (Microsoft AI, Ignite 2026)](https://microsoft.ai/news/microsoft-ignite-2026/)

Source: AI/TLDR LLM registry, https://ai-tldr.dev/models/mai-cyber-2-flash/
