Google DeepMind unveils Gemini 4 Carbon with top benchmark scores and restricted preview

Google DeepMind announced a new artificial intelligence system called Gemini 4 Carbon, according to an update published on the company blog. The new model captured the top position on the Artificial Analysis Intelligence Index, a benchmark that measures technical capabilities across major industry products. The system is specifically engineered to handle complicated software workflows that require dozens of sequential actions.

For regular consumers and office workers, this development points toward a future where digital assistants handle tedious computer projects autonomously. Instead of just answering simple questions or writing short emails, future software tools based on this engine could potentially construct entire applications, fix software bugs, or manage multi step administrative projects without needing constant guidance.

Google is taking a cautious approach to the rollout by keeping the technology away from the general public for now. Access is currently limited to a select group of cybersecurity defenders, allowing specialists to evaluate potential risks and strengthen defenses before wider release.

Giving defensive experts a head start makes sense from a safety perspective, but it raises questions about how long these restricted preview periods should last. It remains unclear how Google will determine when the system is safe enough for general commercial use, or whether rival developers will adopt similar gatekeeping practices for their own high performing tools.