Claude API beta launches Message Threads: long conversations no longer resend history each turn, faster replies, same price

Anthropic has launched Message Threads in public beta for its Messages API: the server holds the conversation for a few minutes, so the next request only needs the new message and the previous reply ID instead of resending the full history. Official figures show median time to first output down about 35% for requests of 500,000 to 1 million tokens, and median request size down from about 800 KB to 1.8 KB in long conversations. Pricing, token consumption and rate limits are unchanged, and threads are distinct from prompt caching, which still needs its own breakpoints. It matters because long agent sessions get faster and lighter on the wire without changing what developers pay, removing a practical bottleneck for multi-hundred-request conversations.

• 宝玉 (Baoyu)2026-11-02 19:34