Cerebras Systems Powers OpenAI's GPT-5.6 Sol, Achieves Record Inference Speed

Cerebras Systems is reportedly powering OpenAI's GPT-5.6 Sol model, which launched in September 2026 and operates at 750 tokens per second in production. This represents the fastest disclosed inference speed for a frontier AI model, highlighting Cerebras's significant contribution to AI infrastructure. The company's wafer-scale engine (WSE-3) within its CS-3 system is designed to accelerate large AI model training and inference by minimizing networking overhead.