StepFun_ai has launched its new model, Step 3.7 Flash, which is designed for enhanced agent efficiency, effectively supporting workflows in coding, searching, and multimodal tasks. This release emphasizes balanced performance, offering 400 TPS, a 198B sparse mixture of experts, and the ability to understand various media formats like UIs and images, ultimately writing code or executing tasks based on its analysis. The model aligns with the current trend in AI to combine speed and cost efficiency while ensuring reliable execution, functioning seamlessly with tools such as Claude Code and Hermes Agent.

StepFun AI: StepFun AI develops AI models optimized for agentic, coding, search, and multimodal tasks. The organization released the Step 3.7 Flash model to advance efficiency in agent workflows while providing open weights. This positions the company as a contributor to practical AI agent tooling.
Hermes Agent: Hermes Agent is an agent framework referenced in discussions around new AI model releases. It is highlighted as a system that pairs with the Step 3.7 Flash model for potential performance evaluation. User interest focuses on compatibility and results when using it with the new release.
Step 3.7 Flash: Step 3.7 Flash is the latest model from StepFun AI built specifically for agent efficiency across coding, search, and multimodal applications. It supports integration with existing agent systems and emphasizes reliable tool use without drift. The release makes the model available via multiple platforms including GitHub and Hugging Face.

Model Focus: Recent AI releases target balanced performance in agentic workflows combining speed, cost efficiency, and dependable tool execution.
Integration Emphasis: New models are designed to work seamlessly with existing agent tools and protocols for multimodal understanding and action.