Google DeepMind Unveils Gemini 4 Carbon for Select Cybersecurity Experts
Google DeepMind has introduced Gemini 4 Carbon, a new AI model that, according to benchmark results, makes progress across all areas. Access is initially restricted to selected security experts due to advanced cyber defense capabilities.
Google DeepMind has introduced Gemini 4 Carbon, a new AI model that, according to the Artificial Analysis Intelligence Index v5, leads this comparison with a score of 59 points. The model appears only two weeks after Gemini 4 Argon, its predecessor from the same model series. According to the company, Gemini 4 Carbon improves performance across all measured areas.
Google describes the model as specifically developed for extended agent workflows and the entire lifecycle of cyber defense. Agent workflows refer to AI systems that independently plan and execute multi-step tasks rather than only answering individual queries. In the field of cybersecurity, the model is meant not only to detect individual threats but also to support security teams across multiple phases of an incident, from detection to response.
Precisely because of these expanded security capabilities, access is initially limited. Gemini 4 Carbon is for now available only to selected security experts. The restriction reflects a widespread concern in the industry: powerful AI tools can help defenders but could equally be used by attackers if access were broad. The company did not state, in the present reporting, when or to what extent a larger user group will receive access.
The short sequence of the two models shows how quickly major AI developers now release new versions. Only two weeks lie between Argon and Carbon, which underscores the pressure in the competition for the top position in performance comparisons. The Artificial Analysis Intelligence Index serves as one of the visible benchmarks against which models can be measured. However, independent detailed evaluations of individual tasks are not available in the current body of sources.
Open questions remain regarding technical details of architecture and training methods, as well as information on pricing and the timeline for general availability. Whether the restriction to security experts will be relaxed in the future also cannot be derived from the available information.
7
