StepFun

StepFun (also known as Stepfun) is a Shanghai-based AI lab founded in 2023, focused on open-weights multimodal foundation models. The lab is best known for its Step model series — Step-1, Step-2, Step-Audio, the widely-adopted Step 3.5 Flash, and the current flagship Step 3.7 Flash — released under the Apache 2.0 license alongside hosted inference on the StepFun Open Platform, OpenRouter, and NVIDIA NIM.

StepFun's design philosophy emphasizes practical agentic deployment: a sparse mixture-of-experts architecture that keeps active parameters small relative to total parameters, native vision-language capabilities, and strong tool-use reliability for coding and search workflows. The lab targets the open-weights tier alongside DeepSeek, Moonshot AI's Kimi, and Liquid AI rather than competing with frontier US labs on raw capability ceiling.

To broaden reach, StepFun partners with hosted inference providers — OpenRouter, NVIDIA NIM, DeepInfra, Fireworks AI, and Modal — so customers can consume Step models without self-hosting while still being able to download the open weights for full control. The lab's headline 2026 release, Step 3.7 Flash, is a 198 billion total parameter mixture-of-experts vision-language model with roughly 11 billion active parameters per token, a 256,000-token context window, and reported throughput up to 400 tokens per second.