StepFun opens Step 5 Preview at $1 per million

StepFun opened Step 5 Preview, a 600B-parameter sparse mixture-of-experts model with 27B active parameters and a 1M-token context window. StepFun’s model documentation lists the context and multimodal inputs, while its pricing table sets cache-miss input at $1 and output at $2.70 per million tokens. That gives builders another cheap long-context candidate to benchmark against Chinese models already winning American workloads, but output-token reasoning costs still need a workload-level trial.