{"slug": "stepfun-publishes-600b-model-weights-with-nvidia-gb300-serving-support", "title": "StepFun publishes 600B model weights with NVIDIA GB300 serving support", "summary": "StepFun published the weights for Step 5 Preview, a 600-billion-parameter model with 27 billion active parameters per token, on Hugging Face on October 15th, following its September 20th introduction. NVIDIA said the NVFP4 version runs at nearly 8,000 tokens per second per GPU on its GB300 platform and called it StepFun's largest open-weights release, though NVIDIA's 24-hour framing remains a company claim rather than independently demonstrated performance.", "body_md": "# StepFun publishes 600B model weights with NVIDIA GB300 serving support\n\n**Step 5 Preview has 600 billion total parameters and 27 billion active per token. NVIDIA says its NVFP4 version runs at nearly 8,000 tokens per second per GPU on GB300.**\n\n        By [Ryan Merket](https://runtimewire.com/author/ryan-merket)\n        · Published \n\nPrimary source: [NVIDIA AI](https://x.com/NVIDIAAI/status/2110730497907216588)\n\n## Why it matters\n\nDownloadable weights and NVIDIA's serving claims give developers a way to test Step 5 outside StepFun's API. The model card describes long-horizon tool use, while NVIDIA's 24-hour framing remains a company claim, not independently demonstrated performance.\n\nStepFun's 600-billion-parameter [Step 5 Preview](https://www.stepfun.com/step-5-preview?ref=runtimewire) weights are available on [Hugging Face](https://huggingface.co/stepfun-ai/Step-5-Preview-NVFP4?ref=runtimewire). NVIDIA says the NVFP4 version runs at nearly 8,000 tokens per second per GPU on its GB300 platform. StepFun introduced the model on September 20th and said weights would follow on October 15th; NVIDIA described the release as StepFun's largest open-weights release in its [October 15th post](https://x.com/NVIDIAAI/status/2110730497907216588?ref=runtimewire).…", "url": "https://wpnews.pro/news/stepfun-publishes-600b-model-weights-with-nvidia-gb300-serving-support", "canonical_source": "https://runtimewire.com/article/stepfun-step-5-preview-open-weights-nvidia-gb300", "published_at": "2026-10-15 16:01:37+00:00", "updated_at": "2026-10-15 19:25:49.820947+00:00", "lang": "en", "topics": ["large-language-models", "ai-infrastructure", "ai-chips", "ai-products"], "entities": ["StepFun", "Step 5 Preview", "NVIDIA", "GB300", "Hugging Face", "NVFP4"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/stepfun-publishes-600b-model-weights-with-nvidia-gb300-serving-support", "markdown": "https://wpnews.pro/news/stepfun-publishes-600b-model-weights-with-nvidia-gb300-serving-support.md", "text": "https://wpnews.pro/news/stepfun-publishes-600b-model-weights-with-nvidia-gb300-serving-support.txt", "jsonld": "https://wpnews.pro/news/stepfun-publishes-600b-model-weights-with-nvidia-gb300-serving-support.jsonld"}}