SSyncDev

MAI-Cyber-1.1-Flash

SyncDev model directory · Profile updated Nov 17, 2026

Compare with another model

MAI-Cyber-1.1-Flash is Microsoft AI's vulnerability-discovery model, available only inside its MDASH scanning harness to verified security customers.

Key specifications

Specification score
4.9
License
Proprietary

The specification score summarizes recorded features and limits. It is not a benchmark result.

Pricing conditions: No public price; billed as Azure consumption inside MDASH for verified customers.

Inputs and outputs

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

  • Text input

Produces

  • Text output

What is MAI-Cyber-1.1-Flash?

There is no endpoint to call and no per-token price to compare. MAI-Cyber-1.1-Flash runs only inside MDASH, Microsoft's multi-model agentic scanning harness for codebases, and only for verified customers. If you are evaluating it, you are really evaluating MDASH.

Microsoft introduced it on 27 October 2026 as an update to MAI-Cyber-1-Flash, promising sharper findings with fewer tokens.

Who can get access

Access is restricted by design:

  • Selected, verified customers only, enabled through a private preview after an application and review.
  • Defensive security use inside MDASH. Activity outside that scope can trigger alerts and, if it continues, loss of access.
  • No standalone API, so no public price list. Usage is billed as Azure consumption within MDASH, based on tokens.

That rules it out for anyone hoping to bolt a security model onto their own agent framework.

What changed from 1-Flash

The biggest shift is calibration. The first version was tuned security-first, refusing more readily; 1.1 is described as cyber-permissive, refusing less so it can support vulnerability discovery, red-teaming and penetration testing for the customers allowed to use it.

Cost is the other headline. On Microsoft's internal MDASH scanning suite, 1.1 matched 1-Flash's results at about 18% lower cost per scan. That is an average on Microsoft's benchmark, not a promise that your invoice drops by the same share; a scan of a monorepo with unusual languages may behave differently.

Reported architecture

Early reporting describes 1.1 as a fine-tune of 1-Flash and inherits its sparse mixture-of-experts design: 137B total parameters, 5B active, and a 256K context. Microsoft's own public material that we could read does not restate those figures, so treat them as provisional. Modalities are not stated anywhere we found; text in and out is the reasonable assumption for code scanning.

Reported standalone scores include 41.68% on CyberGym end-to-end and 23.0% on WebExploitBench. These measure the model alone and cannot be compared with whole-harness MDASH results.

A date mismatch

The predecessor's debut is dated inconsistently: 27 July 2026 circulates, while Microsoft itself dates its "Introducing MAI Cyber 1 Flash inside MDASH" item to 13 August. If version history matters to your compliance records, ask your Microsoft account team for the official release dates.

Deciding whether to apply

Ask your security team one question first: would you send private source code to a managed scanning service at all? If yes, request a pilot and give MDASH a repository with known, already-fixed vulnerabilities. Count true findings, false positives and token spend per scan, and compare against your current static analysis. That tells you more than any leaderboard.

Microsoft's other recent models include MAI-Code-1.1-Flash and MAI-Thinking-1; the full list is on the Microsoft provider page.

How to evaluate MAI-Cyber-1.1-Flash

Compare the model on the work you need to complete.

Tasks and inputs

MAI-Cyber-1.1-Flash is listed for Text → Text, Coding & Code Generation. It accepts text and produces text. Use the publisher’s input format and preprocessing for your chosen deployment.

Capacity and cost

Measure a completed task using the prompts, documents and tools your application will use. Include reasoning and retries in the result.

Choosing a deployment

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.

Frequently asked questions

No. It runs only inside Microsoft's MDASH scanning harness, for verified customers admitted to a private preview.

These records share a directory category. Check the source card, output type and deployment requirements before treating a model as a replacement.

Google

Gemini 4 Hydrogen

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.

1.05M context

OpenAI

GPT-6 Cyber

GPT-6 Cyber is OpenAI's restricted security model for authorised vulnerability investigation, reproduction and fix verification, for Daybreak Red customers.

1.05M context

particle-ai

Particle-1

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.

1.05M context

recursive

Flint

Recursive Flint is an open-weight 39.8B MoE coding-agent model with 2.5B active parameters, free to try via OpenCode and OpenRouter.

262.14K context

Deepreinforce

Ornith-1.5 122B

Ornith-1.5 122B is an open MIT-licensed MoE coding and agent model with a 262K context, reasoning by default and single-80GB-GPU serving.

262.14K context

MiniMax

MiniMax M3.1 Flash

MiniMax M3.1 Flash is MiniMax's GA open-weight multimodal MoE for coding agents: 428B total, 23B active, 1M context, from $0.15 per 1M input tokens.

1M context