What is Gemini 3.5 Flash Cyber? Google's gated security model

Gemini 3.5 Flash Cyber is a security-focused version of Google’s Flash model, tuned to find and fix software vulnerabilities. It shipped on July 21, 2026, alongside two other Flash-tier models. The key limitation: you probably cannot use it yet. It is a limited-access pilot for governments and trusted partners, with no public API or public pricing.

Try Apidog today

Treat this as an explainer, not a setup guide. There is currently no API key, public model ID, or runnable Cyber code sample. What you can do is understand where the model fits, why Google has restricted access, and how to implement practical alternatives today.

What is Gemini 3.5 Flash Cyber?

Gemini 3.5 Flash Cyber is a security-vulnerability specialist. It is part of Google’s CodeMender effort, which focuses on finding and fixing vulnerabilities in code.

The general Flash model is designed for broad coding and reasoning tasks. Cyber is tuned for a narrower workflow:

1. Inspect code for security flaws.
2. Identify risky patterns or vulnerable logic.
3. Propose patches to remediate the issue.

Google announced it in the same Gemini models post that covered the wider Flash refresh. For the general Flash tier, see the DeepMind Flash model page.

Keep these names separate:

• CodeMender is the security program.
• Gemini 3.5 Flash Cyber is the model tuned for the program’s vulnerability-finding and remediation work.

Also, the model name is 3.5 Flash Cyber, not 3.6. The versioning is intentional and reflects separate model tracks.

The catch: you probably can’t use it yet

Gemini 3.5 Flash Cyber is a limited-access pilot. Google has made it available only to governments and trusted partners. It is not publicly available.

In practical terms:

• No public API: You cannot add a Cyber model ID to your application and call it as you would a general Gemini model.
• No public pricing: Google has not published per-token rates because there is no open access to price. Compare this with the Gemini API pricing documentation for standard Flash models.
• No self-serve signup: There is no console setting or developer-account toggle for Cyber.

If you find a tutorial with a Cyber API endpoint, model string, or pricing table, treat it as unsupported. Google has not released those details for public use.

Why Google gated it

The reason is dual use.

A model that can reliably find vulnerabilities can help defenders patch systems. The same capability can also help attackers discover flaws to exploit. Releasing a strong vulnerability finder broadly creates risk beyond normal coding assistance.

Google is therefore starting with a narrow group of vetted users. A limited rollout lets the model support defensive work while Google evaluates how it is used and how it behaves in real environments.

For developers, the implementation takeaway is simple:

Build your security workflow around tools and models you can access today. Treat Cyber as unavailable unless Google explicitly grants your organization access.

There is no committed public launch date, and public availability may never look like a normal self-service model release.

Where it fits in the Flash family

The July 21 refresh shipped three Flash-tier models:

Gemini 3.6 Flash

Gemini 3.6 Flash is the general-purpose workhorse. It supports broad coding and reasoning tasks, includes a 1M-token context window, has a free tier in Google AI Studio, and is publicly priced at $1.50 per million input tokens and $7.50 per million output tokens.

See what is Gemini 3.6 Flash for the full breakdown.

Gemini 3.5 Flash-Lite

Gemini 3.5 Flash-Lite is the lower-cost, high-volume option. It is the fastest of the three, at around 350 output tokens per second, with pricing of $0.30 per million input tokens and $2.50 per million output tokens.

See what is Gemini 3.5 Flash-Lite for details.

Gemini 3.5 Flash Cyber

Gemini 3.5 Flash Cyber is the security specialist. It is available only to governments and trusted partners, with no public API and no public price.

The version numbers are confusing but important:

• The main workhorse moved to 3.6.
• Flash-Lite remains 3.5.
• Cyber also remains 3.5.

These models are on different version tracks. If you are choosing between the public options, see Gemini 3.6 Flash vs 3.5 Flash.

What developers can use today instead

You cannot run Cyber, but you can implement a practical security-review workflow with public tools.

1. Use Gemini 3.6 Flash for first-pass code reviews

Gemini 3.6 Flash is public and can help identify common risky patterns. Use it as a review assistant, not as a replacement for a security audit.

For example, provide a focused prompt with code and explicit review criteria:

Review this Express.js route for security issues.

Check specifically for:
- Missing authentication or authorization
- Unsafe input handling
- IDOR risks
- Missing rate limits
- Sensitive data exposure
- Incorrect HTTP status codes

Return findings with:
1. Severity
2. Affected line or code block
3. Why it is risky
4. A minimal patch

For an API route such as this:

app.get("/users/:id", async (req, res) => {
  const user = await db.users.findById(req.params.id);
  res.json(user);
});

A useful review should flag that the route has no visible authentication or authorization check. Your remediation may look like:

app.get("/users/:id", requireAuth, async (req, res) => {
  if (req.user.id !== req.params.id && !req.user.isAdmin) {
    return res.status(403).json({ error: "Forbidden" });
  }

  const user = await db.users.findById(req.params.id);

  if (!user) {
    return res.status(404).json({ error: "Not found" });
  }

  res.json(user);
});

This does not turn a general model into a specialized vulnerability scanner. It gives you a repeatable first-pass review process for code you own.

2. Test the API behaviors attackers actually probe

Most API weaknesses are not exotic. Common failures include:

• Missing authentication
• Broken authorization
• Weak transport security
• Contract regressions after deployment

You can test these without a gated model.

An API client such as Apidog can help you store requests, define assertions, and run them repeatedly.

3. Add authentication test cases

For each protected endpoint, create at least three requests:

1. No token
2. Expired or invalid token
3. Valid token

Your expected behavior should be explicit:

For example, a basic test assertion should ensure an unauthenticated request does not succeed:

pm.test("Protected endpoint rejects missing token", () => {
  pm.response.to.have.status(401);
});

A 200 OK response where you expected a 401 Unauthorized is a real security finding.

4. Verify mTLS and client-certificate enforcement

If your service requires client certificates, test both the success and failure paths:

• A request with the expected client certificate should complete successfully.
• A plain request without the certificate should be rejected.

Follow this walkthrough for testing APIs with client certificates and mTLS in Apidog.

5. Schedule contract tests

Save your critical API requests and assertions, then run them on a schedule. This catches regressions when an endpoint changes behavior after a deployment.

Useful assertions include:

pm.test("Status code is successful", () => {
  pm.response.to.have.status(200);
});

pm.test("Response includes required fields", () => {
  const body = pm.response.json();

  pm.expect(body).to.have.property("id");
  pm.expect(body).to.have.property("email");
});

You can schedule recurring API tests in Apidog so failures show up shortly after a regression lands rather than during an incident.

None of this requires a gated model. It requires repeatable tests and the discipline to run them for every deployment. To get started, download Apidog and begin with your authentication cases.

FAQ

Can I access Gemini 3.5 Flash Cyber?

No. It is a limited-access pilot for governments and trusted partners. It is not available to the general public, and there is no self-service way to enable it.

How do I request access?

Access is invite- and partner-based, not open signup. There is no public form that grants access. If your organization is a government body or an established Google security partner, the route runs through your Google relationship.

Everyone else should monitor Google’s official channels, including the Google blog and DeepMind model pages, for status changes.

Is there an API or published price?

Not publicly. There is no callable model ID for general use and no public per-token rate.

What is CodeMender?

CodeMender is Google’s effort to find and fix vulnerabilities in code. Gemini 3.5 Flash Cyber is the model tuned for that work.

What should I use for security work in the meantime?

Use public Gemini 3.6 Flash for code-review prompts, then run standard API security checks against your own services:

• Authentication and authorization tests
• mTLS and client-certificate tests
• Contract and regression tests
• Scheduled checks for critical endpoints

The short version

Gemini 3.5 Flash Cyber is a real model with a narrow role: finding and fixing security vulnerabilities as part of Google’s CodeMender effort.

It is also gated. Governments and trusted partners can use it; general developers cannot. There is no public API, public pricing, or self-service signup.

The actionable path today is to use Gemini 3.6 Flash for general code and security-review prompts, then harden your APIs with automated authentication, mTLS, and contract tests. Keep an eye on Google’s official channels if the access model changes.