launchAnthropic2026-11-12
Anthropic Adds Beta Effort Control to Claude API for Mid-Conversation Adjustment
Anthropic introduced a beta effort control for the Claude API, letting developers insert an empty system message into the sequence to change the reasoning effort level mid-conversation while keeping prompt caching intact.
Anthropic added a beta effort parameter to the Claude API. Developers insert an empty system message and set the target effort level through the new output configuration field, which takes effect from the next user message with new input while prompt caching still hits.
ClaudeAPI推理控制