Step-3.5-Flash

Step-3.5-Flash, released by StepFun in February 2026, is a Mixture-of-Experts model using fine-grained routing to maximize computational efficiency — activating only a small fraction of its total parameters per token to enable fast, cost-effective inference without sacrificing capability on general language and reasoning tasks. It targets developers and organizations that need capable AI at high throughput, reflecting StepFun's focus on efficient model architectures as a competitive differentiator.