FrontierThe story, in brief
StepFun Releases Step 3.7 Flash: A 198B MoE Vision-Language Model for Coding Agents and Search Workflows
198B parameters. Native vision. 256k context. StepFun's Step 3.7 Flash is built for agents—not chat.
The KeyNews take
Why it matters
StepFun enters the vision-language model competition with a MoE architecture optimized for agentic workflows and coding tasks, signaling intensifying competition in multimodal reasoning models designed for enterprise automation.
The key facts
6 to knowStep 3.7 Flash: 198B MoE parameters
Native vision-language capability
256k context window
Advisor Mode feature
Positioned for coding agents and search workflows
Published May 29, 2026
Go to the source
MarkTechPostmarktechpost.com
Publisher excerpt: StepFun releases Step 3.7 Flash, a 198B MoE model with native vision, 256k context, and Advisor Mode.