StepFun releases Step 3.7 Flash, an efficient Flash model for production-grade agents
AI summary
StepFun released and open-sourced Step 3.7 Flash, using a sparse MoE architecture (196B+1.8B total parameters, 11B activated), with a maximum generation speed of 400 Tokens/s. It is optimized around four capabilities: native multimodal understanding and execution, enhanced web and visual search, highly reliable tool calling and orchestration, and agent ecosystem compatibility. It reaches 49.5% on Toolathlon, 67.1% on ClawEval-1.1, 45.8% on GDPval, and over 98% pass rate on τ²-bench Telecom. It is compatible with mainstream architectures such as Claude Code and KiloCode as well as the MCP/Skills protocols, supports cloud and local deployment, and has completed integration validation in ecosystems including Kilo Code.
Key Takeaways
- StepFun released and open-sourced Step 3.
- 7 Flash,using a sparse MoE architecture (196B+1.
- 8B total parameters,11B activated),with a maximum generation speed of 400 Tokens/s.
- It is optimized around four capabilities: native multimodal understanding and execution,enhanced…
- It reaches 49.
Quick Facts
- Published
- Updated
- Source
- 公众号:阶跃星辰(Step)
- Category
- Model Releases
- AI score
- 61
Step 3.7 Flash 用激活仅 11B 的 MoE 架构把 Agent 工作流稳定性做透了,兼容主流框架还开源,对需要低延迟、高可靠性的生产环境 Agent 是真正可用的选择。
Related entities
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: StepFun releases Step 3.7 Flash, an efficient Flash model for production-grade agents. May 29, 2026. https://new.nat6.net/en/news/cmq2o63pf01dqsl6nl8z2lff7/