Matthew Berman discusses Google’s announcement of Gemini 4 Carbon, describing the reported benchmarkA benchmark is a defined test or collection of tasks used to compare a system's performance under specified conditions. results and the planned staged rollout. He notes that he had not yet used the model when recording.
Matthew Berman compares reported coding, computer-useComputer use lets an AI agent interact with software through interface controls such as screens, buttons, keyboards and pointers., science and security benchmark scores with other models. He also discusses Google’s stated effort to reduce shortcuts and unsupported claims.
Matthew Berman explains how token usage and pricing contribute to cost per taskCost per completed task measures the total AI, tool, infrastructure, retry, and repair expense for each verified useful outcome., and compares the reported efficiency of models using an intelligence-versus-cost chart. His projections and preferences are presented as his own views.
Watch on YouTube
