[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-post-\u002Fnews\u002Fanthropic-message-threads-claude-code-faster-long-sessions":3,"news-related-en-\u002Fnews\u002Fanthropic-message-threads-claude-code-faster-long-sessions":112,"hreflang-anthropic-message-threads-2026-11":132},{"doc":4,"redirect":8},{"id":5,"title":6,"author":7,"author_avatar":8,"body":9,"category":92,"date":93,"description":94,"extension":95,"image":96,"meta":97,"navigation":98,"path":99,"seo":100,"seoDescription":101,"seoTitle":102,"stem":103,"tags":104,"translationKey":110,"updatedAt":8,"__hash__":111},"news_en\u002Fnews\u002Fanthropic-message-threads-claude-code-faster-long-sessions.md","Anthropic message threads cut the wait on long Claude Code sessions","Paul Lefizelier",null,{"type":10,"value":11,"toc":85},"minimark",[12,30,39,44,57,60,64,67,70,73,77],[13,14,15,16,20,21,29],"p",{},"On 2 November 2026, Anthropic's platform release notes added message threads to the Messages API, in beta, under the header ",[17,18,19],"code",{},"message-threads-2026-08-12",". ",[22,23,28],"a",{"href":24,"rel":25,"target":27},"https:\u002F\u002Fplatform.claude.com\u002Fdocs\u002Fen\u002Frelease-notes\u002Foverview",[26],"nofollow","_blank","The notes"," say a long conversation can now send a smaller request and get a lower time to first token, because the API keeps the conversation on the server and later requests send only the new messages.",[13,31,32,33,38],{},"The product post, ",[22,34,37],{"href":35,"rel":36,"target":27},"https:\u002F\u002Fclaude.com\u002Fresources\u002Farticles\u002Fmessage-threads",[26],"Make long-running agents faster with message threads",", says the feature is on by default for most Claude Code users. On Claude Code requests of 500,000 to 1 million tokens, median time to first token drops by about 35%.",[40,41,43],"h2",{"id":42},"what-actually-changes","What actually changes",[13,45,46,47,52,53,56],{},"Without a thread, every turn of a coding agent uploads the whole conversation again, including earlier tool results. The ",[22,48,51],{"href":49,"rel":50,"target":27},"https:\u002F\u002Fplatform.claude.com\u002Fdocs\u002Fen\u002Fbuild-with-claude\u002Fmessage-threads",[26],"docs"," say the API then validates and tokenizes all of it again. With a thread, the first request sends the full conversation, the system prompt, and the tools, and sets ",[17,54,55],{},"\"thread\": {\"type\": \"create\"}",". Each later request sends only the new messages and the id of the previous response. System and tools stay as they were on the create call.",[13,58,59],{},"The article is explicit about the bill. Each request is smaller, but Claude still reads the full conversation on every turn, so token usage stays the same. Standard token rates and existing rate limits apply. The same beta is on the Claude API, Claude Platform on AWS, Claude in Microsoft Foundry, and Claude Models on Google Cloud when Anthropic manages the models.",[40,61,63],{"id":62},"the-five-minute-limit","The five-minute limit",[13,65,66],{},"A thread is not a saved chat. The article says each thread is held in memory, encrypted, and not written to disk. It expires five minutes after its last request. The client then resends the full conversation and starts a new thread. The docs add that the API deletes a thread no more than one hour after its last use, and that a continue call returns a 404 when the thread is gone. Your application is still the source of truth: the docs say no response tells you whether that turn was stored. You find out on the next continue.",[13,68,69],{},"The docs also separate this from session threads in Claude Managed Agents. Same word, different object.",[13,71,72],{},"The post lists where the latency cut matters most: long coding agents, where each tool result or file edit adds to the conversation; computer-use agents, where screenshots stay with the thread instead of being uploaded again; and document or image workflows. The more context the agent already carries, the more the first token moves up.",[40,74,76],{"id":75},"what-this-is-for-people-waiting-on-an-answer","What this is for people waiting on an answer",[13,78,79,80,84],{},"The part a person feels in a long Claude Code session is the pause before the first token. Anthropic's own account is that the pause gets shorter on the biggest requests, and that the invoice does not. ",[22,81,83],{"href":82},"\u002Fadvertisers\u002F","Idlen's advertisers page"," sells a native ad during that wait, in the editor or the browser, on CPM or CPC from €10. This beta does not add a new surface, and it does not change what that page prints.",{"title":86,"searchDepth":87,"depth":87,"links":88},"",2,[89,90,91],{"id":42,"depth":87,"text":43},{"id":62,"depth":87,"text":63},{"id":75,"depth":87,"text":76},"tools","2026-11-03T14:20:00","On 2 November 2026 Anthropic put message threads in public beta. Most Claude Code users get them by default. On requests of 500k to 1M tokens, median time to first token falls about 35%. The token bill does not.","md","\u002Fimages\u002Fnews\u002Fanthropic-message-threads-claude-code-latency.jpg",{},true,"\u002Fnews\u002Fanthropic-message-threads-claude-code-faster-long-sessions",{"title":6,"description":94},"Message threads, in public beta since 2 November, drop median time to first token about 35% on huge Claude Code requests. Token use stays the same.","Anthropic message threads speed long Claude Code sessions","news\u002Fanthropic-message-threads-claude-code-faster-long-sessions",[105,106,107,108,109],"anthropic","claude-code","messages-api","latency","coding-agents","anthropic-message-threads-2026-11","K2BZKeCCPuz5HESKFyqNSprkQOgz8CkTa1pLS3pFNeM",[113,120,126],{"title":114,"description":115,"path":116,"date":117,"image":118,"category":119},"OpenAI says Astra and Codex are now drafting physical products","On 3 November 2026 OpenAI described internal work where models design enclosures, boards, and CAD. Two people built a Codex device with Astra in under a week. A newer internal model scored 0.875 on BenchCAD 2.0, against 0.594 for GPT-6 Astra.","\u002Fnews\u002Fopenai-astra-engineering-physical-products-cad","2026-11-03T16:10:00","\u002Fimages\u002Fnews\u002Fopenai-physical-engineering-astra-cad.jpg","ai",{"title":121,"description":122,"path":123,"date":124,"image":125,"category":119},"Anthropic's IPO filing says it may soon cross its own threshold for AI-accelerated AI research","In its amended S-1, Anthropic added a risk factor saying internal estimates for future models indicate it may \"imminently\" meet its Responsible Scaling Policy's definition of \"dramatically accelerated progress attributable to AI\". No current model has crossed it. Here is what the threshold means and what the filing says about Claude building Claude.","\u002Fnews\u002Fanthropic-s1-recursive-self-improvement-rsp-threshold-ai-rd","2026-11-03T10:20:00","\u002Fimages\u002Fnews\u002Fanthropic-s1-recursive-self-improvement-rsp-threshold-ai-rd.jpg",{"title":127,"description":128,"path":129,"date":130,"image":131,"category":92},"OpenAI rebuilt the Git backend behind its monorepo to keep up with Codex agents","OpenAI has detailed Pando, a Git-compatible backend it built in 10 weeks with Codex. It coordinates each branch separately, reuses packfiles across fetches, cut median push latency by nearly 5x and now serves close to a petabyte of Git data a day.","\u002Fnews\u002Fopenai-pando-git-backend-monorepo-codex-agents","2026-11-03T10:15:00","\u002Fimages\u002Fnews\u002Fopenai-pando-git-backend-monorepo-codex-agents.jpg",{"en":133,"fr":181,"es":8},{"id":5,"title":6,"author":7,"author_avatar":8,"body":134,"category":92,"date":93,"description":94,"extension":95,"image":96,"meta":178,"navigation":98,"path":99,"seo":179,"seoDescription":101,"seoTitle":102,"stem":103,"tags":180,"translationKey":110,"updatedAt":8,"__hash__":111},{"type":10,"value":135,"toc":173},[136,143,148,150,157,159,161,163,165,167,169],[13,137,15,138,20,140,29],{},[17,139,19],{},[22,141,28],{"href":24,"rel":142,"target":27},[26],[13,144,32,145,38],{},[22,146,37],{"href":35,"rel":147,"target":27},[26],[40,149,43],{"id":42},[13,151,46,152,52,155,56],{},[22,153,51],{"href":49,"rel":154,"target":27},[26],[17,156,55],{},[13,158,59],{},[40,160,63],{"id":62},[13,162,66],{},[13,164,69],{},[13,166,72],{},[40,168,76],{"id":75},[13,170,79,171,84],{},[22,172,83],{"href":82},{"title":86,"searchDepth":87,"depth":87,"links":174},[175,176,177],{"id":42,"depth":87,"text":43},{"id":62,"depth":87,"text":63},{"id":75,"depth":87,"text":76},{},{"title":6,"description":94},[105,106,107,108,109],{"id":182,"title":183,"author":7,"author_avatar":8,"body":184,"category":92,"date":93,"description":250,"extension":95,"image":96,"meta":251,"navigation":98,"path":252,"seo":253,"seoDescription":254,"seoTitle":255,"stem":256,"tags":257,"translationKey":110,"updatedAt":8,"__hash__":261},"news_fr\u002Ffr\u002Fnews\u002Fanthropic-fils-de-messages-sessions-longues-claude-code.md","Les fils de messages d'Anthropic raccourcissent l'attente des longues sessions Claude Code",{"type":10,"value":185,"toc":245},[186,196,203,207,218,221,225,228,231,234,238],[13,187,188,189,20,191,195],{},"Le 2 novembre 2026, les notes de version de la plateforme Anthropic ont ajouté les fils de messages à l'API Messages, en bêta, sous l'en-tête ",[17,190,19],{},[22,192,194],{"href":24,"rel":193,"target":27},[26],"Les notes"," disent qu'une longue conversation peut désormais envoyer une requête plus petite et obtenir un temps plus court avant le premier token, parce que l'API garde la conversation côté serveur et que les requêtes suivantes n'envoient que les nouveaux messages.",[13,197,198,199,202],{},"Le billet produit, ",[22,200,37],{"href":35,"rel":201,"target":27},[26],", dit que la fonction est activée par défaut pour la plupart des utilisateurs de Claude Code. Sur les requêtes Claude Code de 500 000 à 1 million de tokens, le temps médian avant le premier token baisse d'environ 35 %.",[40,204,206],{"id":205},"ce-qui-change-vraiment","Ce qui change vraiment",[13,208,209,210,214,215,217],{},"Sans fil, chaque tour d'un agent de code renvoie toute la conversation, y compris les résultats d'outils déjà vus. ",[22,211,213],{"href":49,"rel":212,"target":27},[26],"La documentation"," dit que l'API valide et tokenise alors tout une seconde fois. Avec un fil, la première requête envoie la conversation complète, le prompt système et les outils, et pose ",[17,216,55],{},". Chaque requête suivante n'envoie que les nouveaux messages et l'identifiant de la réponse précédente. Le système et les outils restent ceux de la création.",[13,219,220],{},"Le billet est net sur la facture. Chaque requête est plus petite, mais Claude relit toute la conversation à chaque tour, donc l'usage en tokens ne change pas. Les tarifs habituels et les limites de débit existantes s'appliquent. La même bêta est sur l'API Claude, Claude Platform sur AWS, Claude dans Microsoft Foundry, et Claude Models sur Google Cloud quand Anthropic gère les modèles.",[40,222,224],{"id":223},"la-limite-de-cinq-minutes","La limite de cinq minutes",[13,226,227],{},"Un fil n'est pas une conversation enregistrée. Le billet dit que chaque fil est gardé en mémoire, chiffré, et non écrit sur disque. Il expire cinq minutes après sa dernière requête. Le client renvoie alors toute la conversation et ouvre un nouveau fil. La documentation ajoute que l'API supprime un fil au plus une heure après sa dernière utilisation, et qu'un appel de continuation renvoie une 404 quand le fil a disparu. L'application reste la source de vérité : la documentation dit qu'aucune réponse n'indique si ce tour a été stocké. On le découvre à la continuation suivante.",[13,229,230],{},"La documentation sépare aussi cela des fils de session de Claude Managed Agents. Même mot, autre objet.",[13,232,233],{},"Le billet liste où le gain de latence compte le plus : les agents de code longs, où chaque résultat d'outil ou édition de fichier allonge la conversation ; les agents de computer use, où les captures restent dans le fil au lieu d'être renvoyées ; et les flux riches en documents ou en images. Plus l'agent porte déjà de contexte, plus le premier token avance.",[40,235,237],{"id":236},"pour-qui-attend-une-réponse","Pour qui attend une réponse",[13,239,240,241,244],{},"Ce qu'une personne sent dans une longue session Claude Code, c'est la pause avant le premier token. Le récit d'Anthropic est que cette pause raccourcit sur les plus grosses requêtes, et que la facture ne baisse pas. La ",[22,242,243],{"href":82},"page annonceurs"," d'Idlen vend une pub native pendant cette attente, dans l'éditeur ou le navigateur, en CPM ou CPC dès 10 €. Cette bêta n'ajoute pas de surface, et elle ne change pas ce que cette page imprime.",{"title":86,"searchDepth":87,"depth":87,"links":246},[247,248,249],{"id":205,"depth":87,"text":206},{"id":223,"depth":87,"text":224},{"id":236,"depth":87,"text":237},"Le 2 novembre 2026, Anthropic a mis les fils de messages en bêta publique. La plupart des utilisateurs de Claude Code les ont par défaut. Sur les requêtes de 500 000 à 1 million de tokens, le temps médian avant le premier token baisse d'environ 35 %. La facture en tokens, non.",{},"\u002Ffr\u002Fnews\u002Fanthropic-fils-de-messages-sessions-longues-claude-code",{"title":183,"description":250},"En bêta publique depuis le 2 novembre, les fils de messages baissent d'environ 35 % le temps médian avant le premier token sur les très grosses requêtes Claude Code. L'usage en tokens ne change pas.","Les fils de messages accélèrent les longues sessions Claude Code","fr\u002Fnews\u002Fanthropic-fils-de-messages-sessions-longues-claude-code",[105,106,258,259,260],"api-messages","latence","agents-de-code","jDLSmNS7V9Q7wiNSH5_bRzKNPjoUyR405If_uvjsrq8"]