Quick answer: Google announced Gemini 4 Carbon on October 28, 2026, positioning it as its strongest model for cybersecurity and long, multi-step tasks. It currently leads the independently run Artificial Analysis Intelligence Index v5.0 with a score of 59, one point ahead of the next model shown. But Carbon is not broadly available: Google says initial access is for a subset of cyber defenders, with no public launch date yet.

What Google says Gemini 4 Carbon is built to do

Google describes Carbon as a frontier AI model for complex coding, agent workflows and cyber defense. The company says its internal teams have used it on sustained engineering work, such as tracing subtle bugs across large codebases and developing fixes. Those examples are Google’s account of internal use, not a guarantee of the model’s performance on every company’s production systems. Google’s announcement highlights cybersecurity as the model’s main differentiator.

What does its AI leaderboard result mean?

On Artificial Analysis’ current Intelligence Index v5.0, Gemini 4 Carbon (High) scores 59. The next entry, Claude Opus 5.5, scores 58. Artificial Analysis says the index combines nine evaluations that it runs independently, so this is stronger evidence than a score reported only by the model maker. Still, a one-point lead on a composite index does not mean Carbon wins every task or will be the best fit for a particular company. See Artificial Analysis’ live index and methodology.

Google also says Carbon leads 15 of the 17 benchmark rows in its own comparison table. Independent coverage points out that rival models remain ahead on some individual software-engineering and computer-use tests. The results are best read as a sign of progress, not a universal ranking of which assistant will perform best on every real-world job. Mashable’s breakdown and VentureBeat’s analysis describe the benchmark trade-offs and access limits.

Google’s cybersecurity claims, with context

Google reports scores of 88.2% on SEC-bench Pro for finding vulnerabilities, 54.8% on ExploitGym for validating exploits, and 71.0% on CWE-bench for patching. These figures come from the company’s announcement. Google notes that the external cybersecurity benchmarks were evaluated without guardrails. The scores are benchmark results, not evidence that Carbon will find every flaw, prevent a breach or safely make production changes without human review.

Who can use Gemini 4 Carbon now?

At launch, Google said it would begin making Carbon available to a subset of cyber defenders and then expand access to additional trusted partners through its Fairwind Program. Developers, businesses and consumers are expected to get access through Google products later, but Google has not announced a date. VentureBeat reports that Carbon does not yet have a publicly available API, so most developers cannot simply start building with it today.

What will Gemini 4 Carbon cost?

Google says the planned introductory API price is $2 per million input tokens and $10 per million output tokens. Cached input tokens are listed at 95% below the input-token price. After the introductory period, Google says the rates will be $4 per million input tokens and $20 per million output tokens. The announcement does not say when that introductory period will end. Because the API is not publicly available yet, these are announced future rates, not a price most teams can currently use to buy access.

The takeaway for businesses and developers

Gemini 4 Carbon gives Google a new benchmark lead and puts cyber defense at the center of its latest model launch. The practical test is still ahead: broader access, independent evaluation on real workflows, clear technical documentation and evidence about how well the model performs with human oversight. For now, the leaderboard result is notable, but the model remains a controlled preview rather than a generally available tool.

Frequently asked questions

Is Gemini 4 Carbon available to the public?

No. Google says access is starting with a subset of cyber defenders and trusted partners. A general developer, business or consumer release date has not been announced.

Does Gemini 4 Carbon outperform every other AI model?

No. It has the highest score on Artificial Analysis’ v5.0 composite index at 59, one point ahead of the next model listed. That does not mean it leads every benchmark or every type of task.

How much will Gemini 4 Carbon cost?

Google announced planned introductory rates of $2 per million input tokens and $10 per million output tokens, followed by $4 and $20 after the introductory period. The model is not yet available through a public API, and Google has not stated when the introductory rates will end.