StepFun showcases Step 5 Preview inference performance on NVIDIA GB300
Original title:Great to see Step 5 Preview delivering strong inference performance on NVIDIA GB300.
AI summary
StepFun says Step 5 Preview delivers strong inference performance on NVIDIA GB300 and looks forward to applications developers build with it. The NVIDIA content it shared calls the model StepFun's largest open-weight release, with a 600B MoE architecture that can handle agent tasks lasting up to 24 hours, reaching nearly 8K tokens/sec/GPU on GB300 with NVFP4, offering an NVFP4 checkpoint (https://huggingface.co/stepfun-ai/Step-5-Preview-NVFP4), and supporting SGLang, vLLM, NeMo Automodel post-training and one-click deployment with NVIDIA NIM.
Key Takeaways
- StepFun says Step 5 Preview delivers strong inference performance on NVIDIA GB300 and looks…
- The NVIDIA content it shared calls the model StepFun's largest open-weight release,with a 600B MoE…
- offering an NVFP4 checkpoint,and supporting SGLang,NeMo Automodel post-training and one-click…
Quick Facts
- Published
- Updated
- Source
- X:阶跃星辰 StepFun (@StepFun_ai)
- Category
- Model Releases
- Topic
- Model Releases
- AI score
- 66
原文补充了 Step 5 Preview 在 NVIDIA GB300 上的推理表现和 NVFP4 权重入口,可帮助开发者评估部署选择。
Topics
Related entities
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: StepFun showcases Step 5 Preview inference performance on NVIDIA GB300. October 15, 2026. https://new.nat6.net/en/news/l3g1uzem751ki1oxckrxa1b3e/