We're expanding which refusals are billed to include refusals that arrive before any output when…
By Steven Van ·
Refusals in the bio, frontier_llm, and reasoning_extraction categories are now billed even when they arrive before any output.
Anthropic is billing more refusals than before. Until now, only refusals that arrived mid-stream were charged; a refusal returned before any output was free. That changes for three categories: bio, frontier_llm, and reasoning_extraction, the categories where Anthropic says it measures low volumes of false positives. A refusal in one of these categories is now billed like a normal request, at the rate of the model that ran it, even when it arrives before any output.
Refusals before any output in other categories, or with no category at all, are still free. Fallback credit, the discount applied when a refused request is retried on another model, is unchanged. The change applies on the Claude API, Amazon Bedrock, Claude Platform on AWS, Google Cloud, and Microsoft Foundry. See how refusals are billed.