阶跃Step 3.7 Flash登顶AA榜：速度、性价比、端到端三项第一 – 量子位
Quick Answer
Step 3.7 Flash has topped the AA leaderboard by prioritizing efficiency, speed, and cost-effectiveness.
Quick Take
As AI transitions to enterprise-level agents, the focus has shifted from single-query performance to high-frequency interactions, significantly impacting operational costs for developers and businesses.
Key Points
- Step 3.7 Flash leads in efficiency, speed, and cost-effectiveness.
- The shift from single-query performance to multi-query efficiency is notable.
- High-frequency interactions increase operational costs for developers.
- AI's transition to enterprise agents changes competition dynamics.
- Significant funding and strategic leadership changes are underway at Step.
Source Excerpt
只能说，停停停各位先别跑分了，咋感觉Step 3.7 Flash这次换成效率打法了呢。
# 模型竞赛的赛点，从单次最强挪到了效率优先
其实前两年大模型圈子的画风还比较简单。
大家比的是啥？比谁聪明。
各家埋头刷Benchmark，成绩多拿一两分就能小火一把，当时比拼的核心是模型单次问答能力的上限。
但今年风向变了。
相信大家也都发现，OpenAI、Anthropic、Cursor嘴里念叨的关键词换成了企业级Agent、工具调用、任务交付……随着Agent成为主流落地形态，原先只拼单点性能的老一套竞赛规则不适用了。
在进入真实业务的Agent工作链路后，AI要反复调用工具、多轮检索信息、分步拆解复杂任务。
问题就出在这儿。
一次回答慢两秒，你能忍，但Agent一个任务要调几十次模型，每次都慢两秒，谁的火气都大。
调一次贵一点没感觉，调几十几百次之后，账单突然就开始在你血压上蹦迪了。
这种高频交互带来的调用量暴涨，推理延迟、Token消耗成倍飙升，不光是咱日常用着闹心，开发者和企业的钱包更是遭不住啊。…
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from WebSearch (Tavily)
See more →Stop just chatting with AI. Learn to build production-ready software in ...
The 2026 Bootcamp offers hands-on training in building production-ready software using Generative AI, applications, and AI agents, emphasizing practical skills over casual interaction with AI. Participants will learn to develop applications like Cursor AI, preparing them for real-world challenges in AI development.