English
Español
Google just fired a major shot in the AI race. The company launched Gemini 4 Carbon, calling it its most advanced model yet, and the pitch is aimed squarely at the parts of AI that are starting to matter most to businesses: writing code, defending against cyberattacks, and running long, complex tasks on its own. For OpenAI and Anthropic, who have spent the past year trading the lead on exactly these capabilities, the arrival of a serious third contender from Google changes the math.
The reason this launch lands harder than a typical model update is where Google is strong. Most people already use Google products every day, from Gmail to Docs to Search. If the company can fold a top-tier AI into tools millions of businesses already pay for, it does not have to win customers one at a time. It already has them.
What Gemini 4 Carbon can do
Google is positioning Gemini 4 Carbon as a model built for serious, long-running work rather than quick questions and answers. It is designed to handle complex engineering and coding tasks that unfold over extended periods, the kind of multi-step jobs that used to require a human to stay in the loop the whole way through.
Cybersecurity is the headline use case. Google says the model can find software vulnerabilities, check whether they can actually be exploited, prioritize the most dangerous ones, and even suggest fixes at the root cause. The company reported a score of 88.2 percent on a benchmark for identifying real-world software vulnerabilities, and said it rolled the model out first to cybersecurity defenders before opening it up to developers and businesses more broadly.
On coding, Google claims Carbon produces more secure code and shows lower rates of the kind of errors and fabrications that make AI risky to trust with real work. By the company's own benchmarks, it ranked at the top of a widely watched AI intelligence index. Those are Google's own numbers, so they deserve a healthy dose of caution, but the direction is clear.
Why OpenAI and Anthropic should be nervous
The threat to OpenAI and Anthropic is not just that Google built a strong model. It is that Google can put that model everywhere its users already are. A business that pays for Google Workspace could get these capabilities built right into the tools its team uses all day, with no new vendor to sign up for and no new system to learn.
That distribution advantage is something neither OpenAI nor Anthropic can easily match. They have excellent products, but they still have to convince each customer to come to them. Google starts with billions of users already inside its ecosystem. When the gap in raw model quality narrows, distribution becomes the deciding factor, and that is the game Google plays best.
What it means for your work
For businesses, the practical takeaway is that the AI tools you rely on are about to get more capable, and the competition between the big three is working in your favor. More powerful models that can run long tasks on their own mean more of the repetitive, time-consuming work can be handed off, from reviewing code to catching security holes before they become breaches.
It is worth keeping expectations grounded. The model is rolling out in stages, the weights are closed, and some of the benchmark claims come straight from Google without independent confirmation. The lower launch pricing is also temporary, and the cost is set to double once the introductory period ends, which matters for any business planning to use it at scale.
Still, the bigger story is the shape of the race. A year ago the AI contest looked like a two-horse sprint. Now the company with the most users on earth has a model that can hold its own. For OpenAI, the leader everyone measures against, that is a reason to look over its shoulder. For everyone else, it is a sign that the tools are getting better and cheaper to access, at least for now.
