Cerebras Systems introduced CS-4, a rack-scale AI server for inference workloads, with first systems expected this quarter and company-reported speeds of up to 30 times those of GPU alternatives under certain workloads. Separately, its partnership with AI orchestration startup Callosum integrates Cerebras accelerators into a platform that routes workflow components to suitable models and hardware in real time, supporting enterprise adoption of agentic AI systems designed to perform tasks with limited human intervention. In company-reported production benchmarks for complex financial-services tasks, the collaboration delivered 4x higher performance, 70% lower compute costs and a 10% increase in task success rates versus a single frontier model on standard infrastructure.