Anthropic opens message threads beta, cutting median TTFT about 35%

Top News

If your agent resends its full conversation on every turn, message threads cut that repeated upload. The Claude Platform feature lets you send only new messages after the initial request, reducing request size and time to first token.

Here is what changed:

• Public beta is available on the native Claude Platform, Claude Platform on AWS, Claude in Microsoft Foundry, and Anthropic-managed Claude Models on Google Cloud.
• Anthropic reports that with threads on by default for most Claude Code users, 500k to 1M-token requests see a median TTFT reduction of about 35%.
• Threads stay encrypted in memory, are not written to disk, and expire five minutes after the last request. After expiry, resend the full conversation.

One catch: Claude still reads the complete conversation every turn. Token usage does not fall, and standard token rates and existing rate limits apply.

Try it: Add the thread parameter to Messages API requests. Send the full conversation first, then only new messages and the previous response ID.

Sources: claude.com