Google's New AI Model Targets Hackers: What Makes Gemini 4 Carbon Different?

Google has introduced Gemini 4 Carbon, a new artificial intelligence model designed to improve cybersecurity, detect software vulnerabilities and handle complex coding tasks.

The announcement comes just weeks after Google introduced Gemini 4 Argon. With Carbon, the company is placing a stronger focus on helping security professionals find software weaknesses, check whether those weaknesses can be exploited and develop fixes.

Google says the model has delivered strong results on several cybersecurity benchmarks. However, access is being introduced gradually, starting with selected cyber-defence teams and trusted partners.

So, what is Gemini 4 Carbon, how is it different from other AI models, and can it help protect businesses from hackers?

What Is Gemini 4 Carbon?

Gemini 4 Carbon is Google's latest frontier AI model, developed by Google DeepMind.

A frontier model is an advanced AI system designed to perform difficult tasks involving reasoning, coding and problem-solving.

Unlike AI tools mainly used for writing, answering questions or creating content, Carbon is built to work on demanding engineering problems that may require many steps.

Its main focus areas include:

• Finding security weaknesses in software
• Testing whether vulnerabilities can be exploited
• Helping developers fix software bugs
• Writing more secure code
• Working through complicated coding projects
• Supporting cybersecurity teams with technical investigations

Google says Carbon can continue working on difficult engineering tasks for extended periods, including problems that require examining large codebases and tracing the root cause of bugs.

Why Is Google Focusing On Cybersecurity?

Cyberattacks can cause serious problems for companies, governments and ordinary internet users. A weakness in software may allow an attacker to access data, disrupt a service or gain control of a system.

Finding and fixing these weaknesses can take time, especially when software contains millions of lines of code.

AI models could help security teams review code, identify suspicious behaviour and suggest possible fixes more quickly.

However, AI can also be misused. The same technical capabilities that help defenders understand a vulnerability may also help an attacker explore it.

This creates a challenge for AI developers: they must improve security capabilities while building safeguards against harmful use.

Google says Gemini 4 Carbon is being released gradually while the company continues testing its safety protections.

What Makes Gemini 4 Carbon Different?

The biggest difference is its focus on the complete software-security process rather than just identifying coding errors.

It Can Help Find Software Vulnerabilities

A software vulnerability is a weakness that could allow a system to behave in an unintended or unsafe way.

Google says Carbon achieved 88.2% on SEC-bench Pro, a benchmark designed to assess the discovery of real-world software vulnerabilities.

This result suggests that the model can be useful for finding complex flaws during security testing. However, a benchmark score does not guarantee that it will identify every vulnerability in real-world systems.

It Can Help Test Security Risks

Finding a possible weakness is only one part of cybersecurity. Security professionals also need to understand whether the weakness creates a practical risk.

Google reports that Carbon scored 54.8% on ExploitGym, a benchmark related to validating vulnerabilities through exploitation.

This capability could help authorised security teams determine which weaknesses deserve urgent attention.

It must be used responsibly: testing should take place only on systems that the user or organisation has permission to assess.

It Can Help Developers Fix Bugs

Identifying a security problem is not enough. Developers also need to correct the underlying cause so the same weakness does not return.

Google says Carbon scored 71.0% on CWE-bench, a benchmark focused on fixing software weaknesses.

If such capabilities work reliably in real projects, they could help development teams spend less time tracing bugs and more time improving software security.

It Is Designed For Long, Complex Tasks

Some engineering problems cannot be solved with a single prompt or a short code change.

Developers may need to inspect several files, understand how different components interact, test possible solutions and check whether a fix creates new problems.

Google says Carbon is designed for these longer, multi-step tasks and that its teams have seen it work through difficult engineering problems over extended periods.

The practical benefit will depend on the task, the quality of the available information and human review.

Gemini 4 Carbon vs Gemini 4 Argon

Carbon follows Google's earlier Gemini 4 Argon model, which also focused on advanced reasoning and professional tasks.

Google says Carbon improves performance in cybersecurity and secure coding.

These comparisons are based on Google's published information. Performance can vary across different tasks, and benchmark results do not necessarily predict how a model will perform in every business environment.

How Much Does Gemini 4 Carbon Cost?

Google announced introductory API pricing of:

• Input: $2 per million tokens
• Output: $10 per million tokens

Cached input tokens are priced at a 95% discount compared with the standard input-token price during the introductory period.

Google says the prices will later increase to $4 per million input tokens and $20 per million output tokens after that period.

These are API prices for processing model inputs and outputs. They are not a monthly subscription price for an ordinary consumer app.

The cost of using Carbon for a business will depend on how much data it processes and how often it is used.

Can Everyone Use Gemini 4 Carbon Now?

No. General access is not available yet.

Google says it is initially making Carbon available to a selected group of cybersecurity defenders. Access will gradually expand to additional trusted partners through its Fairwind Program.

The company plans to make the model available more widely to developers, businesses and consumers, but it has not announced a confirmed date for general availability.

Google is also participating in a voluntary US government process for pre-release model testing as it continues working on safeguards.

For now, developers and ordinary users should not assume that Carbon is available through every Gemini product or API account.

Can Gemini 4 Carbon Stop Hackers?

Gemini 4 Carbon could help cybersecurity teams find and fix weaknesses, but it cannot guarantee complete protection against hackers.

Cybersecurity depends on several things working together, including secure software design, regular updates, access controls, monitoring, backups and trained security staff.

An AI model may miss a flaw, suggest an incorrect fix or fail to understand a particular system's risks. Human experts still need to validate important findings and test changes before deployment.

There is also a wider concern: advanced AI tools can potentially be misused. That is why access controls, safety testing and monitoring are important parts of the rollout.

The more realistic expectation is that Carbon could become another tool for security teams—not a replacement for an organisation's entire cybersecurity programme.

What Could Gemini 4 Carbon Mean For Businesses?

If Carbon performs well outside controlled tests, it could help companies in several ways.

Faster security reviews

Development teams may be able to identify potential weaknesses earlier in the software-development process.

Better bug fixing

AI-assisted analysis could help engineers understand why a problem occurred and suggest a correction.

Support for security teams

Cybersecurity professionals could use the model to investigate complicated issues and prioritise vulnerabilities.

Potential time savings

Automating parts of code review and debugging could reduce the time spent on repetitive technical tasks.

These are possible benefits, not guaranteed outcomes. Businesses would need to evaluate accuracy, costs, data privacy and security risks before using the model in production.

Why Does This Launch Matter In The AI Race?

Google's announcement comes as major AI companies compete to build models that can do more than answer questions or generate text.

Advanced coding and cybersecurity have become important areas of competition because AI tools may help companies develop software, maintain systems and respond to security problems.

Gemini 4 Carbon also follows Gemini 4 Argon by only a short period, showing Google's push to expand its frontier-model lineup.

But benchmark leadership alone does not establish that one model is best for every task. Businesses will also consider reliability, cost, access, integration and safety.