Anthropic's Message Threads cuts time to first token by 35%

Until now, every API request to Claude had to resend the entire conversation history, which meant long exchanges carried a growing payload with each turn. Message Threads changes that by transmitting only what's changed since the last message, and Anthropic says the result is about a 35 percent cut in time to first token for requests between 500,000 and one million tokens. That speedup is aimed at the heaviest users: coding agents, computer-use agents, and workflows processing lots of documents or images, the company says. Notably, the change does not lower token usage, since Claude still reads the full conversation history—so the benefit is latency, not cost. The feature is a public beta on the Claude platform and works through AWS, Microsoft Foundry, and Google Cloud, with threads kept encrypted in memory and expiring after five minutes of inactivity.