[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"portal-settings:stajic:fr":3,"public-menus:all":38,"post:new-qwen-3-5-plus:fr":221,"related:post:new-qwen-3-5-plus:fr:1":466},{"statusCode":4,"data":5,"message":37},200,{"tenantId":6,"lang":7,"defaultLang":8,"siteUrl":9,"contactEmail":10,"brandName":11,"logoUrl":12,"siteName":11,"siteDescription":13,"ogImage":10,"robotsIndex":14,"socialLinks":10,"reservedSlugs":10,"seoPolicy":15},"stajic","fr","de","https:\u002F\u002Fstajic.de",null,"Stajic Platform","\u002FLogo_Planet.svg","Stajic Portal",true,{"branding":16,"relatedContent":17,"crossDomainLinks":18},{"logoUrl":12},{"enabled":14},[19,22,25,28,31,34],{"url":20,"label":21,"isActive":14,"showInFooter":14,"includeInSameAs":14},"https:\u002F\u002Ffigure.rocks","figure.rocks",{"url":23,"label":24,"isActive":14,"showInFooter":14,"includeInSameAs":14},"https:\u002F\u002Floving.rocks","loving.rocks",{"url":26,"label":27,"isActive":14,"showInFooter":14,"includeInSameAs":14},"https:\u002F\u002Fbazify.com","bazify.com",{"url":29,"label":30,"isActive":14,"showInFooter":14,"includeInSameAs":14},"https:\u002F\u002Fbazify.de","bazify.de",{"url":32,"label":33,"isActive":14,"showInFooter":14,"includeInSameAs":14},"https:\u002F\u002Fbazify.at","bazify.at",{"url":35,"label":36,"isActive":14,"showInFooter":14,"includeInSameAs":14},"https:\u002F\u002Fbazify.ba","bazify.ba","Portal settings resolved",[39,45],{"id":40,"name":41,"location":42,"isActive":14,"isDefault":43,"items":44},1,"main-navigation","header",false,[],{"id":46,"name":47,"location":48,"isActive":14,"isDefault":14,"items":49},4,"main-menu","sidebar",[50,66,79,93,103,118,133],{"id":51,"title":52,"url":60,"target":61,"icon":62,"isActive":14,"type":63,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":64,"portfolioId":10,"children":65},"item-18",{"de":53,"en":54,"es":55,"fr":56,"it":54,"ru":57,"sr":58,"zh":59},"Startseite","Home","Inicio","Accueil","Главная","Почетна","首页","\u002Ffull-stack-web-developer-munich-performance-seo-and-maintainable-builds","_self","i-lucide-home","page",111,[],{"id":67,"title":68,"url":75,"target":61,"icon":76,"isActive":14,"type":63,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":77,"portfolioId":10,"children":78},"item-22",{"de":69,"en":69,"es":70,"fr":69,"it":71,"ru":72,"sr":73,"zh":74},"Vision","Visión","Visione","Видение","Визија","想象","\u002Fueber-uns-webdesign-muenchen-webaplikation","i-lucide-eye",113,[],{"id":80,"title":81,"url":89,"target":61,"icon":90,"isActive":14,"type":63,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":91,"portfolioId":10,"children":92},"item-19",{"de":82,"en":83,"es":84,"fr":83,"it":85,"ru":86,"sr":87,"zh":88},"Leistungen","Services","Servicios","Servizi","Услуги","Услуге","服务","\u002Fservices-dienstleistungen-muenchen","i-lucide-wrench",116,[],{"id":94,"title":95,"url":99,"target":61,"icon":100,"isActive":14,"type":63,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":101,"portfolioId":10,"children":102},"item-23",{"de":96,"en":96,"es":96,"fr":96,"it":96,"ru":97,"sr":97,"zh":98},"Blog","Блог","博客","\u002Fblog","i-lucide-book-open",112,[],{"id":104,"title":105,"url":114,"target":61,"icon":115,"isActive":14,"type":63,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":116,"portfolioId":10,"children":117},"item-32",{"de":106,"en":107,"es":108,"fr":109,"it":110,"ru":111,"sr":112,"zh":113},"Neue Technologien","New Technologies","Nuevas tecnologías","Nouvelles technologies","Nuove tecnologie","Новые технологии","Нове технологије","新技术！","\u002Fneue-webtechnologien","i-lucide-sparkles",122,[],{"id":119,"title":120,"url":129,"target":61,"icon":130,"isActive":14,"type":63,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":131,"portfolioId":10,"children":132},"item-20",{"de":121,"en":122,"es":123,"fr":124,"it":125,"ru":126,"sr":127,"zh":128},"Kontakt","Contact us!","Contacto","Contact","Contatto","Контакт","Контактирајте нас","联系我们！","\u002Fcontact","i-lucide-mail",115,[],{"id":134,"title":135,"url":144,"target":61,"icon":145,"isActive":14,"type":63,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":146,"portfolioId":10,"children":147},"item-21",{"de":136,"en":137,"es":138,"fr":139,"it":140,"ru":141,"sr":142,"zh":143},"Unsere Arbeit","Our Work","Nuestro trabajo","Nos réalisations","I nostri lavori","Наши работы","Наши радови","文件夹","\u002Fportfolio","i-lucide-briefcase",114,[148,161,175,191,206],{"id":149,"title":150,"url":144,"target":61,"icon":159,"isActive":14,"type":63,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":146,"portfolioId":10,"children":160},"item-24",{"de":151,"en":152,"es":153,"fr":154,"it":155,"ru":156,"sr":157,"zh":158},"Alle Projekte","All Projects","Todos los proyectos","Tous les projets","Tutti i progetti","Все проекты","Сви пројекти","所有项目","i-lucide-grid-3x3",[],{"id":162,"title":163,"url":171,"target":61,"icon":172,"isActive":14,"type":173,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":10,"portfolioId":10,"children":174},"item-29",{"de":164,"en":165,"es":166,"fr":167,"it":168,"ru":169,"sr":170,"zh":143},"Local Roots, Global Reach","Local Roots - Global Reach","Empresa local ","Entreprise locale","Azienda locale","Местная компания","Локално предузеће глобално тржиште","\u002Fportfolio\u002Flocal-roots-global-reach-communication-media-systems-for-modern-business","i-lucide-globe-2","custom",[],{"id":176,"title":177,"url":186,"target":61,"icon":187,"isActive":14,"type":188,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":10,"portfolioId":189,"children":190},"menu-2916d060-8681-482a-af9a-c107a8bc30d8",{"de":178,"en":179,"es":180,"fr":181,"it":182,"ru":183,"sr":184,"zh":185},"Von Rohdaten zu verifizierbarem KI-Kontext: Eine praktische Architektur für agentenbereite Informationen","From Raw Data to Verifiable AI Context:","De datos sin procesar a contexto de IA verificable: una arquitectura práctica para información lista para agentes","Des données brutes au contexte d'IA vérifiable : une architecture pratique pour des informations prêtes pour les agents","Dai dati grezzi a un contesto IA verificabile: Un'architettura pratica per informazioni pronte per gli agenti","От необработанных данных к верифицируемому контексту ИИ: практическая архитектура информации, готовой для агентов","Od sirovih podataka do proverljivog AI konteksta: Praktična arhitektura za informacije spremne za agente","从原始数据到可验证的 AI 上下文：面向智能体就绪信息的实用架构","\u002Fportfolio\u002Ffrom-raw-data-to-verifiable-ai-context-a-practical-architecture-for-agent-ready-information","i-lucide-database","portfolio",22,[],{"id":192,"title":193,"url":202,"target":61,"icon":203,"isActive":14,"type":188,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":10,"portfolioId":204,"children":205},"menu-e2061cd7-ff53-49d5-a0e8-3d49090577c0",{"de":194,"en":195,"es":196,"fr":197,"it":198,"ru":199,"sr":200,"zh":201},"Präsenzplattformen — Unternehmenskommunikation","Presence Platforms — Enterprise Communication","Plataformas de presencia — Comunicación empresarial","Plateformes de présence — Communication d'entreprise","Presence Platforms — Comunicazione aziendale","Платформы присутствия — Корпоративные коммуникации","Presence platforme — Poslovna komunikacija","Presence 平台 — 企业通信","\u002Fportfolio\u002Fpresence-platforms-enterprise-communication-that-wins-without-competing","i-lucide-badge-percent",21,[],{"id":207,"title":208,"url":217,"target":61,"icon":218,"isActive":14,"type":188,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":10,"portfolioId":219,"children":220},"menu-a781582e-e90d-418e-b8c9-b070bc37c37e",{"de":209,"en":210,"es":211,"fr":212,"it":213,"ru":214,"sr":215,"zh":216},"Maximierung des POS-ROI durch Konsistenz","Maximizing POS ROI Through Consistent","Maximizar el ROI del POS a través de constante","Maximiser le ROI du PDV grâce à la cohérence","Massimizzare il ROI del POS attraverso costante","Максимизация ROI POS за счет последовательного","Maksimizacija POS ROI-ja kroz dosledno","通过持续最大化 POS 投资回报率","\u002Fportfolio\u002Ffrom-vision-to-value-maximizing-pos-roi-through-consistent-ci-creation-and-strategic-branding","i-lucide-briefcase-business",11,[],{"statusCode":4,"data":222,"message":465},{"id":223,"title":224,"slug":225,"content":226,"contentJson":227,"excerpt":319,"featuredImage":320,"featuredImageAlt":321,"featuredImageCaption":10,"featuredImageTitle":10,"featuredImageCopyright":10,"featuredImageAuthor":10,"featuredImageSourceUrl":10,"featuredImageLicense":10,"featuredImageIsAiGenerated":43,"status":322,"publishedAt":323,"createdAt":324,"updatedAt":325,"seoLocalePaths":326,"categories":335,"author":352,"translations":357},"384","Nouveau Qwen 3.5-Plus : l'IA open-source passe aux choses sérieuses","new-qwen-3-5-plus","\u003Ch2>Qwen 3.5-Plus : une IA « agentique » open-source qui réduit les frictions pour les développeurs sur les tâches complexes\u003C\u002Fh2>\n\u003Cp>Alibaba a lancé \u003Cb>Qwen 3.5\u003C\u002Fb> et a particulièrement mis en avant \u003Cb>Qwen 3.5-Plus\u003C\u002Fb> comme un modèle conçu pour le travail \u003Ci>agentique\u003C\u002Fi> : planification, utilisation d'outils et exécution de tâches multi-étapes avec une efficacité nettement accrue. Le message est clair : moins de « magie de prompt », plus d'exécution fiable — le tout avec un contexte allant jusqu'à \u003Cb>1M de tokens\u003C\u002Fb> sur la variante Plus.\u003C\u002Fp>\n\u003Ch3>Pourquoi est-ce important pour les développeurs\u003C\u002Fh3>\n\u003Cp>Si vous construisez des agents de production (RAG, copilotes, bots de revue de code automatiques, pipelines de données, testeurs d'interface utilisateur), le plus gros problème n'est pas de savoir si « le modèle sait quelque chose », mais : \u003Cb>peut-il exécuter un workflow de manière cohérente\u003C\u002Fb> sans s'effondrer à la 6ème étape. Qwen 3.5-Plus cible précisément cette zone — avec un contexte étendu, une entrée multimodale et un comportement d'utilisation d'outils intégré.\u003C\u002Fp>\n\u003Cblockquote class=\"border-l-4 border-gray-300 pl-4 italic\">C'est un modèle qui tente de transformer le LLM d'une « interface de chat » en une \u003Cb>couche d'exécution\u003C\u002Fb> : il voit, planifie, utilise des outils et termine le travail.\u003Ccite class=\"block mt-2 text-sm\">— Comment Qwen 3.5 positionne l'orientation « agentique »\u003C\u002Fcite>\u003C\u002Fblockquote>\n\u003Ch3>Principales nouveautés (Qwen 3.5-Plus en pratique)\u003C\u002Fh3>\n\u003Cul>\u003Cli>\u003Cb>Contexte de 1M\u003C\u002Fb> : cela signifie concrètement que vous pouvez injecter de larges extraits de codebase, des logs, des spécifications et de longues conversations sans « découpage » (chunking) constant.\u003C\u002Fli>\u003Cli>\u003Cb>Utilisation adaptative d'outils\u003C\u002Fb> : le modèle est entraîné pour décider lui-même quand appeler un outil (recherche, exécution de code, navigateur, fonctions) au lieu de tout faire « de tête ».\u003C\u002Fli>\u003Cli>\u003Cb>Multimodal + « agent visuel »\u003C\u002Fb> : il comprend les images\u002Fdocuments et vise à fonctionner via des applications de bureau\u002Fmobiles (un agent capable de « cliquer » et d'exécuter des étapes).\u003C\u002Fli>\u003Cli>\u003Cb>Efficacité (MoE \u002F architecture)\u003C\u002Fb> : focus sur un débit plus élevé et des coûts réduits ; Alibaba souligne dans ses interventions publiques un coût nettement inférieur et une meilleure mise à l'échelle des charges de travail.\u003C\u002Fli>\u003Cli>\u003Cb>Écosystème ouvert\u003C\u002Fb> : la série est accompagnée de versions open-weight et d'outils (repo, formats HF), tandis que le Plus est souvent proposé comme modèle hébergé pour la latence et la stabilité en production.\u003C\u002Fli>\u003C\u002Ful>\n\u003Cdiv class=\"ce-delimiter cdx-block my-8\">\u003C\u002Fdiv>\n\u003Ch3>Comment l'essayer rapidement (sans fioritures)\u003C\u002Fh3>\n\u003Cp>Le chemin le plus rapide passe par un fournisseur qui héberge déjà Qwen 3.5-Plus (par exemple, une passerelle\u002Fagrégateur ou un studio cloud). Si vous avez déjà une application utilisant une API de style « chat completions », la migration consiste principalement à changer le nom du modèle et à vérifier les limites de contexte et d'outils.\u003C\u002Fp>\n\u003Cpre class=\"code-block\" data-lang=\"javascript\">\u003Ccode class=\"language-javascript\">\u002F\u002F Exemple minimal (pseudo) : remplacez l'endpoint\u002FSDK selon le fournisseur\nimport OpenAI from "openai";\n\nconst client = new OpenAI({ apiKey: process.env.API_KEY, baseURL: process.env.BASE_URL });\n\nconst res = await client.chat.completions.create({\n  model: "qwen3.5-plus",\n  messages: [\n    { role: "system", content: "Vous êtes un agent qui termine des tâches." },\n    { role: "user", content: "Parcourez ce dépôt et suggérez 5 améliorations de sécurité." }\n  ]\n});\n\nconsole.log(res.choices[0].message.content);\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Ch3>Cas d'utilisation où Qwen 3.5-Plus présente un réel avantage\u003C\u002Fh3>\n\u003Col>\u003Cli>\u003Cb>RAG agentique sur de grands corpus\u003C\u002Fb> : le contexte de 1M + l'utilisation d'outils réduisent le besoin de résumé (summarization) agressif.\u003C\u002Fli>\u003Cli>\u003Cb>Codage au niveau du dépôt (repo-level)\u003C\u002Fb> : analyse de plusieurs fichiers + génération de PR avec un plan cohérent (moins de « patchwork aléatoire »).\u003C\u002Fli>\u003Cli>\u003Cb>Automatisation UI\u002FQA\u003C\u002Fb> : entrée multimodale + orientation « agent visuel » pour les tests de bout en bout et la reproduction de bugs à partir de screencasts\u002Fcaptures d'écran.\u003C\u002Fli>\u003Cli>\u003Cb>Analyse Ops\u002Fincidents\u003C\u002Fb> : grandes quantités de logs + exécution de runbooks avec des outils (recherche, requête, ticketing).\u003C\u002Fli>\u003Cli>\u003Cb>Agent de workflow de données\u003C\u002Fb> : génération de SQL, validation des résultats, corrections itératives — le tout en une seule session sans perte de contexte.\u003C\u002Fli>\u003C\u002Fol>\n\u003Ch3>Compromis (pour ne pas tomber dans le piège de la hype)\u003C\u002Fh3>\n\u003Cul>\u003Cli>\u003Cb>Contexte de 1M ≠ 1M de « mémoire parfaite »\u003C\u002Fb> : plus l'entrée est grande, plus vous devez faire attention à la structure (sectionnement, index, plan de récupération).\u003C\u002Fli>\u003Cli>\u003Cb>Le comportement agentique nécessite des garde-fous (guardrails)\u003C\u002Fb> : ajoutez impérativement une couche de politique (outils autorisés, limite de débit, sandbox), de journalisation et de relecture.\u003C\u002Fli>\u003Cli>\u003Cb>Hébergé vs open-weight\u003C\u002Fb> : le Plus en tant que modèle hébergé est excellent pour la latence\u002Fstabilité, mais les variantes open-weight sont préférables pour la confidentialité et le contrôle sur site (on-prem) — avec une charge opérationnelle plus élevée.\u003C\u002Fli>\u003C\u002Ful>\n\u003Cp>\u003Cb>En résumé :\u003C\u002Fb> Qwen 3.5-Plus est le signe que la course passe de « qui est le plus intelligent en chat » à « qui exécute le plus fidèlement des workflows complexes ». Si vous développez des agents en production, cela vaut la peine d'être testé — surtout lorsque vos goulots d'étranglement sont le contexte, l'utilisation d'outils et la stabilité à travers plusieurs étapes.\u003C\u002Fp>",{"time":228,"blocks":229,"version":318},1771515753023,[230,235,240,245,249,256,260,271,275,279,283,289,293,303,307,314],{"id":231,"data":232,"type":42},"qwen35p-hero",{"text":233,"level":234},"Qwen 3.5-Plus : une IA « agentique » open-source qui réduit les frictions pour les développeurs sur les tâches complexes",2,{"id":236,"data":237,"type":239},"qwen35p-lead",{"text":238},"Alibaba a lancé \u003Cb>Qwen 3.5\u003C\u002Fb> et a particulièrement mis en avant \u003Cb>Qwen 3.5-Plus\u003C\u002Fb> comme un modèle conçu pour le travail \u003Ci>agentique\u003C\u002Fi> : planification, utilisation d'outils et exécution de tâches multi-étapes avec une efficacité nettement accrue. Le message est clair : moins de « magie de prompt », plus d'exécution fiable — le tout avec un contexte allant jusqu'à \u003Cb>1M de tokens\u003C\u002Fb> sur la variante Plus.","paragraph",{"id":241,"data":242,"type":42},"qwen35p-why",{"text":243,"level":244},"Pourquoi est-ce important pour les développeurs",3,{"id":246,"data":247,"type":239},"qwen35p-why-p",{"text":248},"Si vous construisez des agents de production (RAG, copilotes, bots de revue de code automatiques, pipelines de données, testeurs d'interface utilisateur), le plus gros problème n'est pas de savoir si « le modèle sait quelque chose », mais : \u003Cb>peut-il exécuter un workflow de manière cohérente\u003C\u002Fb> sans s'effondrer à la 6ème étape. Qwen 3.5-Plus cible précisément cette zone — avec un contexte étendu, une entrée multimodale et un comportement d'utilisation d'outils intégré.",{"id":250,"data":251,"type":255},"qwen35p-quote",{"text":252,"caption":253,"alignment":254},"C'est un modèle qui tente de transformer le LLM d'une « interface de chat » en une \u003Cb>couche d'exécution\u003C\u002Fb> : il voit, planifie, utilise des outils et termine le travail.","Comment Qwen 3.5 positionne l'orientation « agentique »","left","quote",{"id":257,"data":258,"type":42},"qwen35p-features",{"text":259,"level":244},"Principales nouveautés (Qwen 3.5-Plus en pratique)",{"id":261,"data":262,"type":270},"qwen35p-features-list",{"items":263,"style":269},[264,265,266,267,268],"\u003Cb>Contexte de 1M\u003C\u002Fb> : cela signifie concrètement que vous pouvez injecter de larges extraits de codebase, des logs, des spécifications et de longues conversations sans « découpage » (chunking) constant.","\u003Cb>Utilisation adaptative d'outils\u003C\u002Fb> : le modèle est entraîné pour décider lui-même quand appeler un outil (recherche, exécution de code, navigateur, fonctions) au lieu de tout faire « de tête ».","\u003Cb>Multimodal + « agent visuel »\u003C\u002Fb> : il comprend les images\u002Fdocuments et vise à fonctionner via des applications de bureau\u002Fmobiles (un agent capable de « cliquer » et d'exécuter des étapes).","\u003Cb>Efficacité (MoE \u002F architecture)\u003C\u002Fb> : focus sur un débit plus élevé et des coûts réduits ; Alibaba souligne dans ses interventions publiques un coût nettement inférieur et une meilleure mise à l'échelle des charges de travail.","\u003Cb>Écosystème ouvert\u003C\u002Fb> : la série est accompagnée de versions open-weight et d'outils (repo, formats HF), tandis que le Plus est souvent proposé comme modèle hébergé pour la latence et la stabilité en production.","unordered","list",{"id":272,"data":273,"type":274},"qwen35p-delim1",{},"delimiter",{"id":276,"data":277,"type":42},"qwen35p-howto",{"text":278,"level":244},"Comment l'essayer rapidement (sans fioritures)",{"id":280,"data":281,"type":239},"qwen35p-howto-p",{"text":282},"Le chemin le plus rapide passe par un fournisseur qui héberge déjà Qwen 3.5-Plus (par exemple, une passerelle\u002Fagrégateur ou un studio cloud). Si vous avez déjà une application utilisant une API de style « chat completions », la migration consiste principalement à changer le nom du modèle et à vérifier les limites de contexte et d'outils.",{"id":284,"data":285,"type":288},"qwen35p-code1",{"code":286,"language":287},"\u002F\u002F Exemple minimal (pseudo) : remplacez l'endpoint\u002FSDK selon le fournisseur\nimport OpenAI from \"openai\";\n\nconst client = new OpenAI({ apiKey: process.env.API_KEY, baseURL: process.env.BASE_URL });\n\nconst res = await client.chat.completions.create({\n  model: \"qwen3.5-plus\",\n  messages: [\n    { role: \"system\", content: \"Vous êtes un agent qui termine des tâches.\" },\n    { role: \"user\", content: \"Parcourez ce dépôt et suggérez 5 améliorations de sécurité.\" }\n  ]\n});\n\nconsole.log(res.choices[0].message.content);","javascript","code",{"id":290,"data":291,"type":42},"qwen35p-usecases",{"text":292,"level":244},"Cas d'utilisation où Qwen 3.5-Plus présente un réel avantage",{"id":294,"data":295,"type":270},"qwen35p-usecases-list",{"items":296,"style":302},[297,298,299,300,301],"\u003Cb>RAG agentique sur de grands corpus\u003C\u002Fb> : le contexte de 1M + l'utilisation d'outils réduisent le besoin de résumé (summarization) agressif.","\u003Cb>Codage au niveau du dépôt (repo-level)\u003C\u002Fb> : analyse de plusieurs fichiers + génération de PR avec un plan cohérent (moins de « patchwork aléatoire »).","\u003Cb>Automatisation UI\u002FQA\u003C\u002Fb> : entrée multimodale + orientation « agent visuel » pour les tests de bout en bout et la reproduction de bugs à partir de screencasts\u002Fcaptures d'écran.","\u003Cb>Analyse Ops\u002Fincidents\u003C\u002Fb> : grandes quantités de logs + exécution de runbooks avec des outils (recherche, requête, ticketing).","\u003Cb>Agent de workflow de données\u003C\u002Fb> : génération de SQL, validation des résultats, corrections itératives — le tout en une seule session sans perte de contexte.","ordered",{"id":304,"data":305,"type":42},"qwen35p-tradeoffs",{"text":306,"level":244},"Compromis (pour ne pas tomber dans le piège de la hype)",{"id":308,"data":309,"type":270},"qwen35p-tradeoffs-list",{"items":310,"style":269},[311,312,313],"\u003Cb>Contexte de 1M ≠ 1M de « mémoire parfaite »\u003C\u002Fb> : plus l'entrée est grande, plus vous devez faire attention à la structure (sectionnement, index, plan de récupération).","\u003Cb>Le comportement agentique nécessite des garde-fous (guardrails)\u003C\u002Fb> : ajoutez impérativement une couche de politique (outils autorisés, limite de débit, sandbox), de journalisation et de relecture.","\u003Cb>Hébergé vs open-weight\u003C\u002Fb> : le Plus en tant que modèle hébergé est excellent pour la latence\u002Fstabilité, mais les variantes open-weight sont préférables pour la confidentialité et le contrôle sur site (on-prem) — avec une charge opérationnelle plus élevée.",{"id":315,"data":316,"type":239},"qwen35p-bottomline",{"text":317},"\u003Cb>En résumé :\u003C\u002Fb> Qwen 3.5-Plus est le signe que la course passe de « qui est le plus intelligent en chat » à « qui exécute le plus fidèlement des workflows complexes ». Si vous développez des agents en production, cela vaut la peine d'être testé — surtout lorsque vos goulots d'étranglement sont le contexte, l'utilisation d'outils et la stabilité à travers plusieurs étapes.","2.31","Découvrez les fonctionnalités et avantages révolutionnaires de Qwen 3.5-Plus d'Alibaba, une IA open-source qui change la donne pour les développeurs.","\u002Fuploads\u002F2026\u002F02\u002Fnew-qwen-3-5-plus-1771515512741-dcbi9p.webp","new-qwen-3-5-plus-1771515512741-dcbi9p","PUBLISHED","2026-02-19T10:23:00.000Z","2026-02-19T15:23:23.973Z","2026-02-20T20:39:04.434Z",{"en":327,"de":328,"sr":329,"es":330,"fr":331,"it":332,"ru":333,"zh":334},"\u002Fblog\u002Fnew-qwen-3-5-plus","\u002Fde\u002Fblog\u002Fnew-qwen-3-5-plus","\u002Fsr\u002Fblog\u002Fnew-qwen-3-5-plus","\u002Fes\u002Fblog\u002Fnew-qwen-3-5-plus","\u002Ffr\u002Fblog\u002Fnew-qwen-3-5-plus","\u002Fit\u002Fblog\u002Fnew-qwen-3-5-plus","\u002Fru\u002Fblog\u002Fnew-qwen-3-5-plus","\u002Fzh\u002Fblog\u002Fnew-qwen-3-5-plus",[336,340,344,348],{"id":337,"name":338,"slug":339},94,"Runbook : Déploiement","release-runbook",{"id":341,"name":342,"slug":343},106,"Runbook : Rollback IA","ai-rollback",{"id":345,"name":346,"slug":347},88,"Versioning (prompts, modèles)","versioning",{"id":349,"name":350,"slug":351},87,"Playbook : LLMOps","llmops",{"id":353,"login":354,"email":355,"displayName":356},"20","rooth8233","","Aleksandar Stajić",[358,427],{"lang":359,"title":360,"content":361,"contentJson":362,"excerpt":426},"en","New Qwen 3.5-Plus: Open-source AI is getting serious now","{\"time\":1771533358356,\"blocks\":[{\"id\":\"qwen35p-hero\",\"data\":{\"text\":\"Qwen 3.5-Plus: Open-source \\\"agentic\\\" AI that removes friction for developers in complex tasks\",\"level\":2},\"type\":\"header\"},{\"id\":\"qwen35p-lead\",\"data\":{\"text\":\"Alibaba has released \u003Cb>Qwen 3.5\u003C\u002Fb> and particularly highlights \u003Cb>Qwen 3.5-Plus\u003C\u002Fb> as a model developed for \u003Ci>agentic\u003C\u002Fi> work: planning, tool use, and execution of multi-step tasks with significantly higher efficiency. The message is clear: less \\\"prompt magic\\\", more reliable execution – and all with a context of up to \u003Cb>1 million tokens\u003C\u002Fb> in the Plus variant.\"},\"type\":\"paragraph\"},{\"id\":\"qwen35p-why\",\"data\":{\"text\":\"Why this is important for developers\",\"level\":3},\"type\":\"header\"},{\"id\":\"qwen35p-why-p\",\"data\":{\"text\":\"When you build production agents (RAG, copilots, automatic code review bots, data pipelines, UI testers), the biggest problem isn't \\\"whether the model knows something\\\", but: \u003Cb>can it consistently process the workflow\u003C\u002Fb> without falling apart at the 6th step. Qwen 3.5-Plus targets exactly this area – with large context, multimodal input, and integrated tool-use behavior.\"},\"type\":\"paragraph\"},{\"id\":\"qwen35p-quote\",\"data\":{\"text\":\"This is a model that attempts to transform LLMs from a \\\"chat UI\\\" into an \u003Cb>execution layer\u003C\u002Fb>: it sees, plans, uses tools, and completes the task.\",\"caption\":\"How Qwen 3.5 positions the \\\"agentic\\\" direction\",\"alignment\":\"left\"},\"type\":\"quote\"},{\"id\":\"qwen35p-features\",\"data\":{\"text\":\"Key innovations (Qwen 3.5-Plus in practice)\",\"level\":3},\"type\":\"header\"},{\"id\":\"qwen35p-features-list\",\"data\":{\"items\":[\"\u003Cb>1M Context\u003C\u002Fb>: In practice, this means you can process large codebase snippets, logs, specifications, and long conversations without constant \\\"chunking\\\".\",\"\u003Cb>Adaptive Tool Use\u003C\u002Fb>: The model is trained to decide for itself when a tool (search, code execution, browser, functions) should be called, instead of doing everything \\\"in its head\\\".\",\"\u003Cb>Multimodal + \\\"Visual Agent\\\"\u003C\u002Fb>: Understands images\u002Fdocuments and aims at working across desktop\u002Fmobile apps (an agent that can \\\"click\\\" and execute steps).\",\"\u003Cb>Efficiency (MoE \u002F Architecture)\u003C\u002Fb>: Focus on higher throughput and lower costs; Alibaba emphasizes significantly lower costs and better scaling of workloads in public appearances.\",\"\u003Cb>Open Ecosystem\u003C\u002Fb>: The series includes open-weight editions and tooling (repo, HF formats), while Plus is often offered as a hosted model for production latency and stability.\"],\"style\":\"unordered\"},\"type\":\"list\"},{\"id\":\"qwen35p-delim1\",\"data\":{},\"type\":\"delimiter\"},{\"id\":\"qwen35p-howto\",\"data\":{\"text\":\"How to try it quickly (without further ado)\",\"level\":3},\"type\":\"header\"},{\"id\":\"qwen35p-howto-p\",\"data\":{\"text\":\"The fastest way is through a provider that already hosts Qwen 3.5-Plus (e.g., gateway\u002Faggregator or cloud studio). If you already have an app that uses a \\\"Chat Completions\\\" style API, migration usually consists only of changing the model name and checking context limits and tools.\"},\"type\":\"paragraph\"},{\"id\":\"qwen35p-code1\",\"data\":{\"code\":\"\u002F\u002F Minimal example (pseudo): replace endpoint\u002FSDK depending on provider\\nimport OpenAI from \\\"openai\\\";\\n\\nconst client = new OpenAI({ apiKey: process.env.API_KEY, baseURL: process.env.BASE_URL });\\n\\nconst res = await client.chat.completions.create({\\n  model: \\\"qwen3.5-plus\\\",\\n  messages: [\\n    { role: \\\"system\\\", content: \\\"You are an agent that completes tasks.\\\" },\\n    { role: \\\"user\\\", content: \\\"Go through this repo and suggest 5 security improvements.\\\" }\\n  ]\\n});\\n\\nconsole.log(res.choices[0].message.content);\",\"language\":\"javascript\"},\"type\":\"code\"},{\"id\":\"qwen35p-usecases\",\"data\":{\"text\":\"Use cases where Qwen 3.5-Plus offers a real advantage\",\"level\":3},\"type\":\"header\"},{\"id\":\"qwen35p-usecases-list\",\"data\":{\"items\":[\"\u003Cb>Agentic RAG for large corpora\u003C\u002Fb>: 1M context + tool-use reduces the need for aggressive summarization.\",\"\u003Cb>Repo-Level Coding\u003C\u002Fb>: Analysis of multiple files + generation of PRs with a consistent plan (less \\\"random patchwork\\\").\",\"\u003Cb>UI\u002FQA Automation\u003C\u002Fb>: Multimodal input + \\\"Visual Agent\\\" alignment for end-to-end tests and reproducing bugs from screencasts\u002Fscreenshots.\",\"\u003Cb>Ops\u002FIncident Analysis\u003C\u002Fb>: Large amounts of logs + runbook execution with tools (search, query, ticketing).\",\"\u003Cb>Data Workflow Agent\u003C\u002Fb>: SQL generation, result validation, iterative corrections – all in one session without context loss.\"],\"style\":\"ordered\"},\"type\":\"list\"},{\"id\":\"qwen35p-tradeoffs\",\"data\":{\"text\":\"Trade-offs (to avoid falling into the hype trap)\",\"level\":3},\"type\":\"header\"},{\"id\":\"qwen35p-tradeoffs-list\",\"data\":{\"items\":[\"\u003Cb>1M Context ≠ 1M \\\"perfect memory\\\"\u003C\u002Fb>: The larger the input, the more you need to pay attention to structure (sectioning, index, retrieval plan).\",\"\u003Cb>Agentic behavior requires guardrails\u003C\u002Fb>: Be sure to add a policy layer (allowed tools, rate limit, sandbox), logging, and replay.\",\"\u003Cb>Hosted vs. Open-Weight\u003C\u002Fb>: Plus as a hosted model is excellent for latency\u002Fstability, but open-weight variants are better for data privacy and on-prem control – with higher operational overhead (Ops).\"],\"style\":\"unordered\"},\"type\":\"list\"},{\"id\":\"qwen35p-bottomline\",\"data\":{\"text\":\"\u003Cb>Conclusion:\u003C\u002Fb> Qwen 3.5-Plus is a signal that the race is shifting from \\\"who is smarter in chat\\\" to \\\"who executes complex workflows more reliably\\\". If you use agents in production, this is worth a test – especially if context, tool-use, and stability across multiple steps are your bottlenecks.\"},\"type\":\"paragraph\"}],\"version\":\"2.31\"}",{"time":363,"blocks":364,"version":318},1771533358356,[365,368,371,374,377,381,384,392,394,397,400,403,406,414,417,423],{"id":231,"data":366,"type":42},{"text":367,"level":234},"Qwen 3.5-Plus: Open-source \"agentic\" AI that removes friction for developers in complex tasks",{"id":236,"data":369,"type":239},{"text":370},"Alibaba has released \u003Cb>Qwen 3.5\u003C\u002Fb> and particularly highlights \u003Cb>Qwen 3.5-Plus\u003C\u002Fb> as a model developed for \u003Ci>agentic\u003C\u002Fi> work: planning, tool use, and execution of multi-step tasks with significantly higher efficiency. The message is clear: less \"prompt magic\", more reliable execution – and all with a context of up to \u003Cb>1 million tokens\u003C\u002Fb> in the Plus variant.",{"id":241,"data":372,"type":42},{"text":373,"level":244},"Why this is important for developers",{"id":246,"data":375,"type":239},{"text":376},"When you build production agents (RAG, copilots, automatic code review bots, data pipelines, UI testers), the biggest problem isn't \"whether the model knows something\", but: \u003Cb>can it consistently process the workflow\u003C\u002Fb> without falling apart at the 6th step. Qwen 3.5-Plus targets exactly this area – with large context, multimodal input, and integrated tool-use behavior.",{"id":250,"data":378,"type":255},{"text":379,"caption":380,"alignment":254},"This is a model that attempts to transform LLMs from a \"chat UI\" into an \u003Cb>execution layer\u003C\u002Fb>: it sees, plans, uses tools, and completes the task.","How Qwen 3.5 positions the \"agentic\" direction",{"id":257,"data":382,"type":42},{"text":383,"level":244},"Key innovations (Qwen 3.5-Plus in practice)",{"id":261,"data":385,"type":270},{"items":386,"style":269},[387,388,389,390,391],"\u003Cb>1M Context\u003C\u002Fb>: In practice, this means you can process large codebase snippets, logs, specifications, and long conversations without constant \"chunking\".","\u003Cb>Adaptive Tool Use\u003C\u002Fb>: The model is trained to decide for itself when a tool (search, code execution, browser, functions) should be called, instead of doing everything \"in its head\".","\u003Cb>Multimodal + \"Visual Agent\"\u003C\u002Fb>: Understands images\u002Fdocuments and aims at working across desktop\u002Fmobile apps (an agent that can \"click\" and execute steps).","\u003Cb>Efficiency (MoE \u002F Architecture)\u003C\u002Fb>: Focus on higher throughput and lower costs; Alibaba emphasizes significantly lower costs and better scaling of workloads in public appearances.","\u003Cb>Open Ecosystem\u003C\u002Fb>: The series includes open-weight editions and tooling (repo, HF formats), while Plus is often offered as a hosted model for production latency and stability.",{"id":272,"data":393,"type":274},{},{"id":276,"data":395,"type":42},{"text":396,"level":244},"How to try it quickly (without further ado)",{"id":280,"data":398,"type":239},{"text":399},"The fastest way is through a provider that already hosts Qwen 3.5-Plus (e.g., gateway\u002Faggregator or cloud studio). If you already have an app that uses a \"Chat Completions\" style API, migration usually consists only of changing the model name and checking context limits and tools.",{"id":284,"data":401,"type":288},{"code":402,"language":287},"\u002F\u002F Minimal example (pseudo): replace endpoint\u002FSDK depending on provider\nimport OpenAI from \"openai\";\n\nconst client = new OpenAI({ apiKey: process.env.API_KEY, baseURL: process.env.BASE_URL });\n\nconst res = await client.chat.completions.create({\n  model: \"qwen3.5-plus\",\n  messages: [\n    { role: \"system\", content: \"You are an agent that completes tasks.\" },\n    { role: \"user\", content: \"Go through this repo and suggest 5 security improvements.\" }\n  ]\n});\n\nconsole.log(res.choices[0].message.content);",{"id":290,"data":404,"type":42},{"text":405,"level":244},"Use cases where Qwen 3.5-Plus offers a real advantage",{"id":294,"data":407,"type":270},{"items":408,"style":302},[409,410,411,412,413],"\u003Cb>Agentic RAG for large corpora\u003C\u002Fb>: 1M context + tool-use reduces the need for aggressive summarization.","\u003Cb>Repo-Level Coding\u003C\u002Fb>: Analysis of multiple files + generation of PRs with a consistent plan (less \"random patchwork\").","\u003Cb>UI\u002FQA Automation\u003C\u002Fb>: Multimodal input + \"Visual Agent\" alignment for end-to-end tests and reproducing bugs from screencasts\u002Fscreenshots.","\u003Cb>Ops\u002FIncident Analysis\u003C\u002Fb>: Large amounts of logs + runbook execution with tools (search, query, ticketing).","\u003Cb>Data Workflow Agent\u003C\u002Fb>: SQL generation, result validation, iterative corrections – all in one session without context loss.",{"id":304,"data":415,"type":42},{"text":416,"level":244},"Trade-offs (to avoid falling into the hype trap)",{"id":308,"data":418,"type":270},{"items":419,"style":269},[420,421,422],"\u003Cb>1M Context ≠ 1M \"perfect memory\"\u003C\u002Fb>: The larger the input, the more you need to pay attention to structure (sectioning, index, retrieval plan).","\u003Cb>Agentic behavior requires guardrails\u003C\u002Fb>: Be sure to add a policy layer (allowed tools, rate limit, sandbox), logging, and replay.","\u003Cb>Hosted vs. Open-Weight\u003C\u002Fb>: Plus as a hosted model is excellent for latency\u002Fstability, but open-weight variants are better for data privacy and on-prem control – with higher operational overhead (Ops).",{"id":315,"data":424,"type":239},{"text":425},"\u003Cb>Conclusion:\u003C\u002Fb> Qwen 3.5-Plus is a signal that the race is shifting from \"who is smarter in chat\" to \"who executes complex workflows more reliably\". If you use agents in production, this is worth a test – especially if context, tool-use, and stability across multiple steps are your bottlenecks.","Discover the groundbreaking features and benefits of Alibaba's Qwen 3.5-Plus, a revolutionary open-source AI for developers.",{"lang":7,"title":224,"content":226,"contentJson":428,"excerpt":319},{"time":228,"blocks":429,"version":318},[430,432,434,436,438,440,442,445,447,449,451,453,455,458,460,463],{"id":231,"data":431,"type":42},{"text":233,"level":234},{"id":236,"data":433,"type":239},{"text":238},{"id":241,"data":435,"type":42},{"text":243,"level":244},{"id":246,"data":437,"type":239},{"text":248},{"id":250,"data":439,"type":255},{"text":252,"caption":253,"alignment":254},{"id":257,"data":441,"type":42},{"text":259,"level":244},{"id":261,"data":443,"type":270},{"items":444,"style":269},[264,265,266,267,268],{"id":272,"data":446,"type":274},{},{"id":276,"data":448,"type":42},{"text":278,"level":244},{"id":280,"data":450,"type":239},{"text":282},{"id":284,"data":452,"type":288},{"code":286,"language":287},{"id":290,"data":454,"type":42},{"text":292,"level":244},{"id":294,"data":456,"type":270},{"items":457,"style":302},[297,298,299,300,301],{"id":304,"data":459,"type":42},{"text":306,"level":244},{"id":308,"data":461,"type":270},{"items":462,"style":269},[311,312,313],{"id":315,"data":464,"type":239},{"text":317},"Post erfolgreich abgerufen",{"items":467,"source":507,"manualIds":508,"manualMatchedIds":509},[468,474,481,486,493,500],{"id":469,"slug":470,"title":471,"excerpt":472,"featuredImage":320,"publishedAt":473},"445","qwen-3-6-in-production-release-runbook-ai-rollback-and-llmops-versioning","Qwen 3.6 en production : Runbook de déploiement, Rollback IA et Versionnage LLMOps","Qwen 3.6 n'est pas seulement une autre mise à jour de modèle. C'est à la fois un événement de déploiement, un scénario de rollback et un problème de versionnage. Cet article explique comment Qwen 3.6 doit être géré en production à travers la discipline LLMOps, la traçabilité des prompts et des modèles, le déploiement contrôlé et une préparation au rollback basée sur des preuves.","2026-05-04T02:49:00.000Z",{"id":475,"slug":476,"title":477,"excerpt":478,"featuredImage":479,"publishedAt":480},"488","what-is-context-engineering-what-the-model-receives-before-it-answers","Qu’est-ce que l’ingénierie du contexte ? Ce que le modèle reçoit avant de répondre","L'ingénierie de contexte conçoit les informations qu'un modèle d'IA reçoit avant l'inférence, y compris les invites, la récupération, la mémoire, l'état de l'application, les résultats d'outils et l'historique des conversations.","\u002Fuploads\u002F2026\u002F10\u002Fwhat-is-context-engineering-what-the-model-receives-before-it-answers-1791480653258-018kcv.webp","2026-10-08T13:29:00.000Z",{"id":482,"slug":483,"title":483,"excerpt":10,"featuredImage":484,"publishedAt":485},"369","git-with-automatic-upload-and-synchronization-to-a-production-server","\u002Fuploads\u002F2024\u002F05\u002Fstep-by-step-guide-illustration-showing-the-process-of-setting-up-Git-with-auto-upload-and-synchronization-to-a-production-server-large.webp","2024-05-28T22:48:00.000Z",{"id":487,"slug":488,"title":489,"excerpt":490,"featuredImage":491,"publishedAt":492},"459","ollama-is-not-the-product-building-production-ready-open-llm-applications","Ollama n'est pas le produit : construire des applications Open-LLM prêtes pour la production","Exécuter un modèle local avec Ollama est facile. Construire une application Open-LLM prête pour la production est plus difficile : cela nécessite du RAG, du contrôle d'accès, de l'abstraction de fournisseur, de l'évaluation, de la journalisation, de la discipline de déploiement et une couche applicative contrôlée autour du modèle.","\u002Fuploads\u002F2026\u002F06\u002Follama-is-not-the-product-building-production-ready-open-llm-applications-1782679361640-h0usqf.webp","2026-06-28T16:39:00.000Z",{"id":494,"slug":495,"title":496,"excerpt":497,"featuredImage":498,"publishedAt":499},"496","why-does-ai-retrieve-some-sources-and-ignore-others-the-retrieval-authority-layer","Pourquoi l'IA récupère-t-elle certaines sources et en ignore-t-elle d'autres ? — La couche d'autorité de récupération","Trouver un document pertinent ne suffit pas. Les systèmes d’IA doivent aussi distinguer les sources actuelles, faisant autorité, applicables et traçables des sources qui semblent simplement similaires. La couche d’autorité de récupération définit cette frontière de sélection des sources.","\u002Fuploads\u002F2026\u002F10\u002Fwhy-does-ai-retrieve-some-sources-and-ignore-others-the-retrieval-authority-layer-1792072410616-0evp1z.webp","2026-10-15T07:51:00.000Z",{"id":501,"slug":502,"title":503,"excerpt":504,"featuredImage":505,"publishedAt":506},"493","mlops-vs-llmops-what-changes-when-the-model-is-an-llm","MLOps vs LLMOps : qu'est-ce qui change lorsque le modèle est un LLM","MLOps exploite les systèmes d'apprentissage automatique ; LLMOps étend ces pratiques aux prompts, au contexte, à la récupération, aux fournisseurs, aux outils, aux évaluations et au comportement d'exécution autour des grands modèles de langage.","\u002Fuploads\u002F2026\u002F10\u002Fmlops-vs-llmops-what-changes-when-the-model-is-an-llm-1791487319869-2v7hxo.webp","2026-10-08T15:20:00.000Z","fallback",[],[]]