引用ZenMux@ZenMuxAI
Excited to support Step 3.7 Flash by @StepFun_ai on ZenMux from day one. 🚀
A sparse MoE vision-language model built for agentic, coding, search, multimodal, and long-context workflows.
⚡ 400 TPS, ~11B active params, 256K context, 3 reasoning levels
👁️ understands UIs, charts, docs, and images — then writes code or calls tools to act on what it sees
🔎 built for deeper web + visual search, with more sources and follow-up
🛠️ reliable tool use: less drift, fewer broken tool calls, 98%+ on τ²-bench across difficulty levels
🧠 works with Claude Code, KiloCode, Hermes Agent, OpenClaw, and MCP-style workflows
🖥️ runs locally on Mac Studio M4 Max, DGX Spark, and AMD AI Max+ 395
Try it on ZenMux now → https://zenmux.ai/stepfun/step-3.7-flash
One API for capable models.
#StepFun #ZenMux #AIAgents #LLM