Gemini 4 Carbon: Benchmarks, Pricing and Limited Access

SAN FRANCISCO / LONDON (BYTEVYTE) - Google announced Gemini 4 Carbon, its next frontier model, on October 28, 2026. Launch access is restricted to a select group of cyber defenders and trusted partners under the Fairwind Program, and general availability for developers and businesses has not yet been scheduled. Google presents Gemini 4 Carbon as the successor to Gemini 4 Argon, which was released about two weeks earlier, with an emphasis on cybersecurity and long-horizon agentic tasks. The lead claims come from Google's own evaluations, so their weight depends on evidence that outside reviewers have not yet been able to test.

What Gemini 4 Carbon reports

Google says Carbon ranks first on the Artificial Analysis Intelligence Index v5.0. On its own evaluations, the model achieves 88.2% on SEC-bench Pro, 54.8% on ExploitGym and 71.0% on CWE-bench, all of which measure vulnerability work. On the WebDev Arena leaderboard it places fourth at 1769 Elo, a gain of 89 points. Google also reports a 0.0% success rate against the indirect prompt injection attacks it evaluated, and it says internal testing showed an 80% drop in high-severity misalignment alerts compared with earlier versions.

The index score gives a sharper comparison than the vendor figures. The index places Carbon at 59, a one-point lead over Claude Opus 5.5 at 58 and two points ahead of GPT-6 Astra at 57. That margin is narrow, and it is measured on a single composite index rather than on the security tasks Google emphasizes.

Pricing and access

Introductory API pricing comes to $2 per million input tokens and $10 per million output tokens. Once that introductory window closes, the rate climbs to $4 and $20, so both input and output cost twice as much. Cached input tokens are billed 95% below the standard input rate, a saving that counts most for workloads that keep reusing the same large inputs across many calls. The available material does not say how long the introductory period lasts or when general availability will start.

Trade-offs and missing evidence

Three gaps stand out. The cybersecurity figures are vendor-reported and available only to a small partner group, so outside teams cannot yet reproduce them. The announcement does not name the comparison model or the sample size behind a Chrome patch claim. The 0.0% prompt injection result covers only the attacks Google evaluated, so it does not establish resistance across all attack types.

For business buyers, the choice is between early access and independent verification. Organizations inside the Fairwind Program can test Carbon before the wider release, while everyone else must rely on published numbers. A one-point index lead rarely decides procurement, so cost per completed security task and reliability under real workloads will carry more weight once pricing steps up to $4 and $20.

Why this matters

Gemini 4 Carbon tests whether a frontier lab can sell security capability while controlling who gets to use it first. Security teams should track the general availability date and the post-introductory price, then run their own vulnerability detection evaluations before committing budgets.

While we strive for accuracy and rigor, ByteVyte can make mistakes. Readers are advised to verify all information independently and conduct due diligence before making decisions based on this reporting. We accept no liability for errors, omissions, or decisions made in reliance on this content.

Sources

Gemini 4 Carbon: our next frontier model

✔Human Verified

Researched and cross-referenced against primary sources by the Bytevyte editorial team. This article was generated with the assistance of artificial intelligence and reviewed by the Bytevyte editorial team.