Anthropic raises the maximum Claude API request size from 32 MB to 128 MB
Anthropic’s official developer account ClaudeDevs posted on October 29, 2026, that the company has raised the maximum request size on the Messages and Token Counting endpoints of the Claude API from 32 MB to 128 MB. The adjustment is mainly aimed at image-intensive requests, such as long-running computer use sessions and large PDF documents, which previously hit the 32 MB limit even when there was still room in the context window.
The PDF support note in the company’s documentation has been updated to state that the first-party Claude API’s maximum request size is 128 MB. Some overview pages still show 32 MB, indicating that the documentation updates may not yet be fully synchronized. This increase is about four times the original limit.
Original source:
Title: We've raised the max request size for Messages and Token Counting on the Claude API from 32 MB to 128 MB... Publisher: ClaudeDevs (@ClaudeDevs) Publication date: 2026-10-29 URL: https://x.com/ClaudeDevs/status/2115924543113998811（引用帖） Source nature: official company developer account original post Verification status: The full post text and the official PDF support documentation (2026-10-29) were actually read for this report Supporting content: The core statement that the maximum request size for the Messages and Token Counting endpoints was raised from 32 MB to 128 MB
ABAB AI Insight
The request size limit and the model context window (token count) are two separate constraints. Previously, image and PDF content often triggered the byte limit before tokens were exhausted because of their larger encoded size. The 128 MB limit allows a single request to carry more visual content and reduces the extra calls caused by splitting files.
This change applies to Anthropic’s first-party Claude API. Partner platforms (such as Amazon Bedrock and Google Cloud) have their own request size limits and may not raise them at the same time. Users still need to pay attention to token billing and rate limits; a higher byte limit does not change the model’s token-based billing.
Metrics to watch include whether the official documentation is fully updated, whether other endpoints (such as the Batch API) are adjusted, and changes in request success rates for large PDF and computer use scenarios in actual use.
ABAB News · Cognition Law
The byte-size limit on API requests and the token limit of the model context window are independent of each other, and visual content often reaches the byte limit first.
Updates to parameters across different official documentation pages may lag, so the latest announcement and the updated documentation for specific features should prevail.