Table of Contents
Google announced two new frontier artificial intelligence models on Wednesday, just two weeks after its previous release, featuring the advanced large language model Gemini 4 Carbon and the world model Genie 4.
Gemini 4 Carbon and Cybersecurity Focus
Gemini 4 Carbon is the company’s most advanced model to date, surpassing the previous Gemini 4 Argon model and competitors like OpenAI’s GPT-6 Astra on benchmarking tests that evaluate overall capabilities. Google is positioning the model specifically for cybersecurity professionals, noting that it excels at detecting and patching vulnerabilities. According to Google, this capability can uncover major issues, similar to recent PHP vulnerabilities that affected a majority of the world’s websites.
Genie 4 World Model and Audio Features
Alongside the large language model, Google introduced Genie 4, a world model capable of simulating real and imagined physical worlds. While traditional chatbots predict the next word in a sentence, world models operate based on the physics of a digital universe, making them useful for robotics and video game design. Genie 4 introduces a new capability by incorporating audio, allowing developers to add background noise and dialogue to their artificial intelligence-generated worlds.
Availability and Public Outlook
Neither Gemini 4 Carbon nor Genie 4 is currently available to the general public. Access to Gemini 4 Carbon is restricted to select cybersecurity experts, and Google declined to answer when public access will open, though past releases suggest it could take one to two weeks. Google stated that it is actively engaging with the US government’s review process, while users can already join the waitlist for Genie 4.
