All capabilities Cost and throughput

Run the whole job overnight at half price

Submit many requests as one asynchronous batch, poll for completion, and pay 50% less than the same work sent one request at a time.

DocsOfficially documentedTestNot tested by benchr Quick · API Docs checked September 1, 2026

Claude APIClaude Opus 5Claude Sonnet 5Claude Opus 4.8+1

Overview

Processes each request in the batch independently and asynchronously, with most batches finishing in under an hour.
Cuts cost by 50% relative to the synchronous path, which is the whole point for evaluations and bulk classification.
Raises throughput for workloads where nobody is waiting on the answer.

01Identify the work that has no human waiting on it: evaluation runs, backfills, bulk tagging, re-processing an archive.
02Submit the batch, then poll the batch status rather than the individual requests.
03Retrieve results only once processing has ended for the whole batch.
04Stack it with prompt caching where the requests share a long prefix.

It is asynchronous by definition - anything user-facing cannot use it.
Web search inside batches is throttled per organisation to protect shared capacity, so search-heavy batches can take longer.
The 50% discount applies to token cost; server tool charges such as web search are priced the same as in regular requests.

benchr has not run this capability. No test has been published.

benchr test runs →

Source-check history

September 1, 2026DocsOfficially documentedFirst logged. The 50% figure and the sub-hour typical completion read from the batch processing guide.