Glonce

Anthropic launches Message Threads beta, cutting Claude time to first token

toolsThe Glonce editorsThe Decoder🔥66Glonce rating66Impact55Novelty60Relevance95Surprise35Players90Credibility75
Anthropic launches Message Threads beta, cutting Claude time to first token

Anthropic has released Message Threads as a public beta on its Claude platform. Until now, every API request had to resend the entire conversation history. With Message Threads, only what has changed since the last message is transmitted.

Anthropic says this cuts time to first token by about 35 percent for requests between 500,000 and one million tokens. That figure is the company's own claim.

According to Anthropic, coding agents, computer-use agents, and workflows that process lots of documents or images benefit the most.

The thread stays encrypted in memory between requests and expires after five minutes of inactivity. Token usage does not change, because Claude still reads the full conversation history. The feature works on the Claude platform and through AWS, Microsoft Foundry, and Google Cloud. The details are in Anthropic's documentation.

Key facts

  • Anthropic released Message Threads as a public beta on the Claude platform; it sends only what changed since the last message instead of the full conversation history.
  • Anthropic says time to first token drops by about 35 percent for requests between 500,000 and one million tokens.
  • The thread stays encrypted in memory between requests and expires after five minutes of inactivity.
  • Token usage is unchanged, since Claude still reads the full conversation history.
  • It works on the Claude platform and through AWS, Microsoft Foundry, and Google Cloud.

Why it matters

Until now, every Claude API request had to resend the entire conversation history. In very long sessions that means sending a lot of repeated data each time. Message Threads sends only the changes since the last message, and Anthropic says that cuts time to first token by about 35 percent for requests of 500,000 to one million tokens. The wait before a reply starts is the part users feel in long agent sessions.

Who it affects

Anthropic names three groups that benefit the most: coding agents, computer-use agents, and workflows that process lots of documents or images. The feature is available on the Claude platform and through AWS, Microsoft Foundry, and Google Cloud, so developers using Claude through those clouds are covered too.

How to use it

The feature is in public beta, and Anthropic points to its documentation for details. The thread expires after five minutes of inactivity. Token usage does not change, because Claude still reads the full conversation history. The article does not say how the feature is enabled or what the API parameters are; it only points to the documentation.

How solid is it

The report comes from The Decoder and relies on Anthropic's statements. The 35 percent figure is Anthropic's own claim, given as about 35 percent for requests between 500,000 and one million tokens. No independent measurement is cited, and no benchmark methodology, model names or baseline latency in seconds are given.

Risks and caveats

The speedup is stated only for requests of 500,000 to one million tokens, and the figure is approximate. No latency figures are given for requests below 500,000 tokens or above one million tokens. The feature is a beta. No pricing or cost change is stated beyond token usage being unchanged. The thread stays encrypted in memory between requests and expires after five minutes of inactivity.