StepFun’s flagship preview for agentic and professional work, with a 92-block narrow-and-deep Transformer interleaving Gated Sliding Window Attention and Sparse GQA full-attention layers; 600B total parameters, 27B active, and a 1M-token context.
StepFun released the weights on 2026-10-15; official Hugging Face repositories provide FP8, BF16, and NVFP4 checkpoints. The model remains in Preview under the StepFun Model License 1.0 with usage conditions.
Only published specifications are shown.
— Step 5 Preview entered product and API preview on StepFun; open weights are not yet released. ↗ — StepFun opened Step 5 Preview weights on the announced date; FP8, BF16, and NVFP4 checkpoints are accessible under its official organization, with the StepFun Model License 1.0. The model remains in Preview. ↗
Step 5 Preview remains in product/API preview alongside its announced weight release; StepFun has also disclosed its deep attention design and long-horizon agent training methods. Open weights do not imply GA.
Step 5 Preview entered product and API preview on StepFun; open weights are not yet released.
View sourceStepFun opened Step 5 Preview weights on the announced date; FP8, BF16, and NVFP4 checkpoints are accessible under its official organization, with the StepFun Model License 1.0. The model remains in Preview.
View sourceSelect a card to view the related model.
Copy this template, add the field, proposed value, and primary source, and share it with the archive maintainer.