Reported News
Anthropic's message threads beta cuts Claude Code time to first token about 35% on long sessions
Anthropic put message threads into public beta on the Messages API on November 2, 2026: after the first request, an agent sends only the new messages and the previous response ID. Anthropic says that with threads on by default for most Claude Code users, "requests with 500k to 1M tokens see median TTFT drop by about 35%", while token usage stays the same.
- Changes reported
- 6
- listed below, each from a source we read
- We tested
- Not tested yet
- reported from sources; we did not run it ourselves
- Gaps we flag
- 3
- things we could not verify
- Sources
- 3
- 2 primary, linked below
What changed
| # | What the source says | Status |
|---|---|---|
| 01 | New thread parameter: the first Messages API request sends the full conversation, system prompt and tools with thread type "create"; later requests send only the new messages plus the previous response ID with thread type "continue", and leave system and tools out. | Reported |
| 02 | Claude Code already uses it: Anthropic says message threads is on by default for most Claude Code users, and that median time to first token for 500K to 1M token requests dropped by about 35%. | Reported |
| 03 | No cost saving on its own: Claude still reads the whole conversation each turn, so token usage is unchanged and standard token rates and rate limits apply; prompt caching is still what lowers cost. | Reported |
| 04 | Short lifetime: a thread expires 5 minutes after its last request, each stored response restarts that timer, and the API deletes it no more than 1 hour after last use. Continuing an expired thread returns a 404, so clients must keep their own copy and replay it. | Reported |
| 05 | Limits: the docs list it as not eligible for zero data retention or HIPAA, on-demand compaction is not available with threads, and token counting requests that include a thread return a 400 error. Every request needs the message-threads-2026-08-12 beta header. | Reported |
| 06 | Where it runs: public beta on the Claude Platform, Claude Platform on AWS, Claude in Microsoft Foundry and Anthropic-managed Claude models on Google Cloud. | Reported |
reported by the sourcemeasured or run by usnot verified
Primary source: Anthropic: Make long-running agents faster with message threads. We are the fast, independent index; the vendor’s own page is the source of record for this release.
In the source’s words
“With message threads on by default for most Claude Code users, requests with 500k to 1M tokens see median TTFT drop by about 35%.”
“While each request is smaller with message threads, Claude still reads the full conversation on every turn, so token usage stays the same.”
“It expires 5 minutes after its last request, and the API deletes it no more than 1 hour after its last use.”
Early verdict
Worth turning on for long-running agents where latency hurts, as long as your client keeps the full history and handles the 404 replay; it will not lower your bill.
Basis: Based on Anthropic's article and docs, which state the latency claim, unchanged token usage and the 5-minute expiry; we have not run it.
What we have not verified
- We did not run message threads or measure time to first token; the 35% figure is Anthropic's own number for Claude Code.
- Anthropic does not say what share of Claude Code users have threads on, beyond "most".
- We did not check whether the speedup holds for conversations under 500K tokens.
This brief is Reported: written from the sources below, not from running the release ourselves.
Sources
- Anthropic: Make long-running agents faster with message threadsprimary source · fetched 2026-11-03
- Claude Platform docs: Message threadsprimary source · fetched 2026-11-03
- The New Stack coveragesource · fetched 2026-11-03
Related
Spotted an error? See our corrections policy or email the address on the contact page. Markdown version for agents: /news/anthropic-message-threads-claude-code.md
Want these briefs in your Google Top Stories? Add Agentic Weekly as a preferred source on Google.
Newsletter
The Weekly Verdict. Free, every Monday.
A free email every Monday: the week's few big AI releases, then 5 short use cases for the new models, each with a prompt you can paste.
Free. Double opt-in. One click to unsubscribe.