多数实验室事后才考虑推理效率，而 Step 3.7 Flash 从设计之初就为推理优化，做智能体应用和视觉语言模型的开发者可以直接试用，感受 400 tok/sec 的流畅体验。
Step 3.7 Flash 是阶跃星辰（StepFun）发布的 198B 稀疏 MoE 视觉语言模型，专为推理效率从头设计。该模型包含 196B 语言骨干和 1.8B 视觉编码器，支持原生多模态理解和行动，可靠工具使用，以及增强的网页和视觉搜索。在真实智能体工作负载下，推理速度可达 400 tok/sec，并采用 Apache 2.0 开源许可。Fireworks AI 已提供在线试用。
Many research labs only consider inference efficiency after the fact. Step 3.7 Flash is a 198B spars...
Many research labs only consider inference efficiency after the fact. Step 3.7 Flash is a 198B sparse MoE VLM designed by @StepFun_ai for inference from the start. 196B language backbone with a 1.8B vision encoder. Built for real-world agent workloads, running at up to 400 tok/sec. Native multimodal understanding and action, reliable tool use, and enhanced web and visual search. Apache 2.0. Try it now → fireworks.ai/models/firewor… 💬 0 🔄 0 ❤️ 2 👀 182 📊 1 ⚡
- 阶跃星辰 Stepfun06-04 15:01原文