# Anthropic adds in-conversation effort adjustment to Claude API

> Original title: Claude API 新增调节 effort 能力：让 AI 按任务使劲，从闲聊转复杂分析可中途“换挡”

- Source: IThome · AI
- Published: 2026-11-12T02:23:54.000Z
- Bad Influence: https://badinfluence.news/items/ifnfltlcesipt4wfmtk2ofcm1
- Original: https://www.ithome.com/1/025/894.htm

## Summary

Anthropic adds a beta feature to the Claude API that lets developers change the effort setting partway through a conversation. Developers set the target level through a new output_config.effort parameter, inserting an empty system message without changing top-level configuration or prompt caching. The new setting applies from the next user message with new input and does not alter content already generated.