StepFun's Step 5 Preview: 600B open-weight MoE hits ~8K tokens/sec/GPU on GB300
StepFun_ai · x · 2026-10-15
StepFun confirmed its Step 5 Preview, the company's largest open-weights release, is delivering strong inference performance on NVIDIA GB300.
Key specs: a 600B MoE model built for agent tasks lasting up to 24 hours, running at nearly 8K tokens/sec/GPU on GB300 with NVFP4. NVIDIA published an NVFP4 checkpoint with deploy paths via SGLang and vLLM, post-training through NeMo Automodel, and one-command serving on NVIDIA NIM.
Related event: StepFun Open-Sources Step 5 Preview: 600B-Parameter MoE with 1M Context(17 posts)→
More from Infra
- Neon Realtime turns any Postgres query into a live subscription, built inside the database — matei_zaharia · 2026-10-16
- Microsoft unveils in-house 3T-class MAI base models: 80B active params, 50T training tokens detailed — SinclairWang1 · 2026-10-16
- Jensen Huang says Nvidia doesn't need to backstop CoreWeave as its cost of capital falls — ayushthakur0 · 2026-10-16
- Zuckerberg: Meta's Muse Spark Open weights out today, runs on Mac Studios — AIatMeta · 2026-10-16
- Microsoft unveils MAI-Base-2: 2.8T-param MoE, 4-12x more pre-training FLOP-efficient than DeepSeek — ChengleiSi · 2026-10-16
- Local RAG pipeline: retrieval sets the ceiling, not the model — Arindam_1729 · 2026-10-16