Gemini 4 Carbon restricted to cybersecurity before gradual rollout
In brief
- Gemini 4 Carbon leads the Artificial Analysis Intelligence Index with 59 points.
- Access is limited to cybersecurity experts and Fairwind Program partners, with no developer timeline.
- Introductory pricing (2/10 $/M tokens) will double to 4/20 $ at an unspecified date.
Google unveiled **Gemini 4 Carbon** on October 28, 2026, calling it its « most powerful » model. According to the **Artificial Analysis Intelligence Index v5**, it tops the ranking with 59 points, ahead of Claude Opus 5.5 and GPT-6 Astra. Yet access remains tightly controlled: first reserved for a select group of cybersecurity experts, then gradually extended to other trusted partners via the **Fairwind Program**, with no timeline for developers.
The model is available via Google’s API, but pricing is unclear. An introductory rate of 2 and 10 dollars per million tokens (0.10 dollar with cache) is set to rise to 4 and 20 dollars at an unspecified date. Google mentions participation in security testing before public release, without detailing the framework. Anthropic and Mistral have adopted similar restrictions for recent models like **Claude Mythos Preview** or a cybersecurity-focused variant of **Mistral Large 4**.
A broad release of Carbon depends on feedback and guardrail adjustments, with no confirmed schedule. Its one-point lead in the Artificial Analysis index is marginal, and current pricing does not reflect stable commercial availability. Cybersecurity players and trusted partners are the first in line, while the public must wait.
What it changes for you
- CybersecurityCybersecurity firms gain early access to a top-tier model, but must comply with strict contractual terms.
This summary gives you the essentials. The full article, its figures and quotes are on ActuIA.
Read the original articleAI-generated summary of the source, checked by rules. How we write · Report an error