{"id":658518,"date":"2026-09-22T11:56:09","date_gmt":"2026-09-22T09:56:09","guid":{"rendered":"https://aivancity.ai/blog/?p=658518"},"modified":"2026-09-22T15:59:32","modified_gmt":"2026-09-22T13:59:32","slug":"qwen3-8-max-codage-agents-ia-performances-limites","status":"publish","type":"post","link":"https://aivancity.ai/blog/qwen3-8-max-codage-agents-ia-performances-limites/","title":{"rendered":"Qwen3.8-Max : ce que le modèle d’Alibaba change réellement pour le codage et les agents IA"},"content":{"rendered":"\n
\n

\n Launched by Alibaba on\n August 3, 2026, Qwen3.8-Max combines a\n Mixture-of-Experts architecture with\n 2,400 billion parameters—of which\n 95 billion are active—with a reported context window\n of one million tokens and\n multimodal and agent-based capabilities.\n

\n\n

\n The published assessments show clear improvement in coding and on\n certain professional tasks, but they do not establish\n overall superiority.\n

\n\n

\n The model's main advantage lies in the combination of\n available resources, enhanced cloud services, and the ability to run long-running tasks.\n

\n
\n\n\n\n
\n\n
\n 01\n
\n\n

\n What Just Happened\n

\n\n
\n\n\n\n

Alibaba Cloud officially unveiled Qwen3.8-Max on August 3, 2026. The company describes it as its largest and most capable Qwen model to date, with 2,400 billion parameters, an expert architecture, a context window of up to one million tokens, and support for text, images, and video as input in its Model Studio service.[1][2]

\n\n\n\n

API offers this model in several regions, including China, Singapore, Germany, the United States, Japan, and Hong Kong. However, the documentation notes that certain features vary by region. For example, web search is available in some environments but not in all initial service deployments.[2]

\n\n\n\n

Since its launch, the situation has evolved. Alibaba has published the weights of the Qwen3.8-2.4T-A95B model on Hugging Face. The official documentation specifies, however, that Qwen3.8-Max is a service version based on this model, enhanced in particular by visual input, integrated tools, a no-reasoning mode, and a context of one million tokens enabled by default.[3] It would therefore be inaccurate to speak of a single product that is entirely identical across the API and the download.

\n\n\n\n
\n\n
\n 02\n
\n\n

\n What's Really New\n

\n\n
\n\n\n\n

The first change relates to the model’s scale. Qwen3.8-2.4T-A95B has a total of 2,400 billion parameters, but only 95 billion are activated to process each token. This Mixture-of-Experts architecture makes it possible to increase overall capacity without mobilizing the entire network at every step. The technical specifications indicate 512 experts, of which ten routed experts and one shared expert are activated.[3]

\n\n\n\n

The second change concerns agent-based positioning. Qwen3.8-Max is not merely presented as a conversational response system. Alibaba intends it for tasks involving coding, research, professional work, and execution across long sequences. In these scenarios, the model plans multiple steps, uses tools, observes the results, and adjusts its course.

\n\n\n\n

The third change lies in the relationship between the open model and the managed service. Text weights can be downloaded under a specific Qwen3.8-Max license, while the most integrated features are provided in the cloud. This distinction creates two different options: greater control and customization with self-hosting, or more out-of-the-box features via the API.

\n\n\n\n\n\n
\n
\n\n
\n \"Executive\n
\n
\n\n
\n\n \"aivancity\"\n\n
\n

Executive MBA in AI & Business Transformation

\n

The MBA Redesigned for the Age of AI. For experienced executives who want to lead the transformation of their organizations. Paris, Nice, and Dubai.

\n
\n 12 months — Part-time\n At least 10 years of experience\n Early bird: 20,000 €\n Paris · Nice · Dubai\n
\n
\n\n
\n \n Learn more about the program →\n \n \n \n 100% in English\n \n
\n\n
\n
\n
\n\n\n\n
\n\n
\n 03\n
\n\n

\n How Qwen3.8-Max Works\n

\n\n
\n\n\n\n

Qwen3.8-Max is based on a hybrid architecture that combines a Mixture-of-Experts mechanism with several types of attention. A router selects the experts deemed relevant for each token. This allows the model to distribute its capabilities across different specializations without activating all 2,400 billion of its parameters at the same time.

\n\n\n\n

The available weight-based version has a native context size of 262,144 tokens, expandable to approximately 1.01 million tokens. In Model Studio, Alibaba reports a total context size of one million tokens, with up to 991,808 input tokens and 131,072 output tokens depending on the parameters used.[2][3] This capacity facilitates the analysis of large software repositories or voluminous documents, but it does not guarantee that all information will be processed with the same level of accuracy.

\n\n\n\n

In an agent-based workflow, the model receives an objective, breaks down the task, calls functions or tools, and then uses the results to continue its work. For coding, this may include reviewing a code repository, generating changes, and running tests. For professional use, the managed version also supports visual and video content, generates structured text, and can perform web searches when permitted by the region.[2]

\n\n\n\n
\n\n
\n 04\n
\n\n

\n Features and Technical Limitations\n

\n\n
\n\n\n\n
\n \n \n \n \n\n \n \n \n\n \n \n \n\n \n \n\n \n \n\n \n \n\n \n \n\n \n \n\n \n \n\n \n \n\n \n \n\n \n \n\n \n \n\n \n \n \n
\n Capacity or Constraint\n \n What You Need to Know\n
\n MoE Architecture\n \n A total of 2,400 billion parameters, of which 95 billion are activated\n per token, according to Alibaba.\n
\n Context\n \n 262,144 tokens natively for the published weights, with a\n possible extension to approximately 1.01 million. The Max service specifies\n 1 million tokens by default.\n
\n Multimodality\n \n Text, images, and videos are input into Model Studio. The\n documented output remains text-based.\n
\n Agents and Tools\n \n Function calls, structured outputs, integrated tools, and\n Web search by region.\n
\n On-premises deployment\n \n Weights are available on Hugging Face, but the model's scale requires\n a very large infrastructure.\n
\n Customization\n \n Fine-tuning the Max service is listed as unsupported\n in the documentation I consulted.\n
\n
\n\n\n\n
\n\n
\n 05\n
\n\n

\n What Benchmarks Actually Tell Us\n

\n\n
\n\n\n\n

The table published by the Qwen team shows a significant improvement over Qwen3.7-Max. Qwen3.8-Max scores 86.6 on Terminal-Bench 2.1, compared to 74.5 for its predecessor; 67.7 on SWE-bench Pro, compared to 60.6; and 93.0 on PaperBench, compared to 64.8.[3] These results support the idea of progress in coding and reproducibility, under the conditions chosen by the team.

\n\n\n\n
\n \n \n \n \n\n \n\n \n\n \n \n \n\n \n \n \n\n \n\n \n\n \n \n\n \n \n\n \n\n \n\n \n \n\n \n \n\n \n\n \n\n \n \n\n \n \n\n \n\n \n\n \n \n\n \n \n\n \n\n \n\n \n \n \n
\n Evaluation\n \n Qwen3.7-Max\n \n Qwen3.8-Max\n \n Read Carefully\n
\n Terminal-Bench 2.1\n \n 74,5\n \n 86,6\n \n A clear improvement, but GPT-5.6 Sol is listed as 88.8 in the same\n table.\n
\n SWE-bench Pro\n \n 60,6\n \n 67,7\n \n An improvement, but a score lower than the 80.0 given to Fable 5.\n
\n DeepSWE 1.1\n \n 21,6\n \n 56,6\n \n A significant improvement, but still lagging behind Fable 5 and GPT-5.6 Sol.\n
\n PaperBench\n \n 64,8\n \n 93,0\n \n Highest score in the published table, to be confirmed by\n independent reproductions.\n
\n IFBench\n \n 79,1\n \n 82,8\n \n Progress in following instructions in the selected protocol.\n
\n
\n\n\n\n

These figures do not allow us to identify a \"best\" model in general terms. Performance varies greatly depending on the task. Protocols, available tools, the number of attempts, the reasoning budget, and the maximum context may differ. An evaluation conducted or compiled by the vendor remains a useful primary source, but it is no substitute for an independent, standardized comparison.

\n\n\n\n

The Arena ranking provides another indicator, based on user preferences. On September 13, 2026, Qwen3.8-Max was ranked 22nd overall in the text ranking, with a score of 1,481 ± 6 and 16,670 votes.[4] This result confirms its competitiveness, but it also shows that the fifth-place ranking announced at launch was not stable over time. Rankings change as new models are introduced, the number of votes increases, and calculation methods evolve.

\n\n\n\n
\n\n
\n 06\n
\n\n

\n A context of one million tokens does not guarantee perfect memory\n

\n\n
\n\n\n\n

A very large context window allows you to process more documents or code in a single session. It reduces the need to artificially split certain corpora and can make it easier to perform tasks that require references scattered throughout a large dataset.

\n\n\n\n

However, input capacity and operational quality are not the same thing. A model may be able to process one million tokens without reproducing every detail with the same reliability, correctly linking distant elements, or maintaining a consistent strategy throughout the sequence. Performance also depends on how the corpus is organized, how the query is formulated, and the associated information retrieval mechanisms.

\n\n\n\n

For a company, the challenge therefore lies less in filling the window than in preparing the data: selecting sources, version control, logical segmentation, access rights, and evaluating responses. A very long timeframe is no substitute for a document architecture or knowledge governance.

\n\n\n\n
\n\n
\n 07\n
\n\n

\n Cost may favor the use of agencies, but it must be calculated based on the entire assignment\n

\n\n
\n\n\n\n

In the Singapore region, which is intended for international use, Alibaba charges a standard rate of $2 per million tokens for incoming transactions and $6 per million for outgoing transactions. In several other global regions, the documentation lists rates of $1.65 for incoming transactions and $4.951 for outgoing transactions. Specific rates apply to the cache and may vary depending on promotions.[2]

\n\n\n\n

These prices may seem attractive for a model in this category, but the actual cost of an agent isn’t limited to a single call. A long task can involve multiple iterations, runs, tool calls, and checks. It is therefore important to measure the cost per successful task, latency, error rate, and human verification work, rather than simply comparing the listed price per million tokens.

\n\n\n\n
\n\n
\n 08\n
\n\n

\n Why This Matters to Organizations\n

\n\n
\n\n\n\n

Qwen3.8-Max expands companies’ options between managed APIs and controlled deployment. The cloud service provides multimodal features, tools, and regional availability. The published models offer greater opportunities for inspection and customization, but their size makes self-hosting difficult and costly. Technical availability, therefore, does not equate to operational accessibility.

\n\n\n\n

For development teams, the ability to work on long-term projects can automate certain phases of diagnosis, generation, testing, and documentation. However, this automation requires isolated environments, minimal permissions, traceability of actions, and human review of changes before they are integrated.

\n\n\n\n

For knowledge-based professions, multimodality and long-term context can support the analysis of reports, contracts, videos, or internal corpora. Organizations must, however, determine what data can be sent, in which region it is processed, how long it is retained, and what contractual obligations apply.

\n\n\n\n
\n\n
\n 09\n
\n\n

\n Issues of Accountability and Governance\n

\n\n
\n\n\n\n

The first issue concerns data. A one-million-token limit makes it easier to send massive volumes of documents, including personal, confidential, or protected information. In Europe, the GDPR continues to apply when personal data is processed. Companies must verify the legal basis, data minimization, any transfers, security, and individuals’ rights. The AI Act does not replace these obligations.

\n\n\n\n

The second challenge concerns agent autonomy. A model capable of using tools or executing code can cause unintended actions, propagate an error, or access a resource outside its scope. Oversight must be designed prior to deployment: sandboxing, a list of authorized tools, human validation for sensitive operations, logging, and a shutdown mechanism.

\n\n\n\n

The third issue concerns licensing and openness. The datasets are available under a specific license, Qwen3.8-Max, which should not automatically be equated with a standard open-source license.[3] Any commercial reuse requires a review of the applicable terms, particularly regarding redistribution, permitted uses, and liabilities.

\n\n\n\n

Finally, the literature reviewed does not provide a complete energy footprint for training or inference. The MoE architecture reduces the proportion of activated parameters, but this is not sufficient to demonstrate an overall environmental benefit. Model size, long contexts, and agentic trajectories can still be computationally very expensive. In the absence of published measurements, this uncertainty must be acknowledged.

\n\n\n\n
\n\n
\n 10\n
\n\n

\n What to Watch for Now\n

\n\n
\n\n\n\n

The true impact of Qwen3.8-Max will depend on whether its results can be independently replicated, on feedback from production use, and on the stability of its performance on long-running tasks. It will also be important to monitor differences between versions, as the documentation already mentions a Qwen3.8-Max-0902 snapshot that improves coding, agent collaboration, and visual understanding.[2]

\n\n\n\n

The model shows that a very large-scale system can combine available computing resources with enhanced cloud services, even if the two offerings are not strictly identical. This hybridization is likely more significant than the raw number of parameters. For organizations, the key question is not whether Qwen3.8-Max outperforms all its competitors, but whether it meets a measurable need with an acceptable level of cost, control, security, and compliance.

\n\n\n\n
\n
\n

\n\n

\n\n

\n Learn more\n

\n\n

\n To learn more about the transformations illustrated by\n Qwen3.8-Max, check out these\n analyses focusing on open-weight models, AI agents,\n technological sovereignty, and data management.\n

\n
\n\n
\n \n
\n \n Open-weight models\n \n\n \n Kimi K3: What You Need to Know About the Open-Source Model with\n 2,800 billion parameters\n \n
\n\n \n Read the analysis →\n \n \n\n \n
\n \n Agent-based AI\n \n\n \n Gemini 3.7 Flash: Google Steps Up Efforts on Code and AI Agents\n \n
\n\n \n Read the analysis →\n \n \n\n \n
\n \n Technological Sovereignty\n \n\n \n After Mistral: Who Will Hold Sovereignty in the New AI World Order?\n \n
\n\n \n Read the analysis →\n \n \n\n \n
\n \n Data and Deployment\n \n\n \n The Blind Spot in AI Training: Data Management\n \n
\n\n \n Read the analysis →\n \n \n
\n
\n\n\n\n
\n
\n

\n Sources\n

\n
\n\n
\n

\n [1]\n Alibaba Cloud, August 3, 2026,\n Alibaba Unveils Qwen3.8-Max: Its Largest and Most Capable Flagship Model to Date.\n \n View source\n \n

\n\n

\n [2]\n Alibaba Cloud Model Studio, documentation\n Qwen3.8-Max, updated September 11, 2026, accessed on\n September 22, 2026.\n \n View the documentation\n \n

\n\n

\n [3]\n Qwen, official page\n Qwen3.8-2.4T-A95B on Hugging Face, accessed on\n September 22, 2026.\n \n View the model page\n \n

\n\n

\n [4]\n Arena Intelligence,\n Text Arena Leaderboard, as of September 13, 2026.\n \n View the leaderboard\n \n

\n\n

\n [5]\n The Verge, August 3, 2026,\n China’s Alibaba Takes Another Swipe at America’s AI Supremacy.\n \n Read the article\n \n

\n
\n
\n","protected":false},"excerpt":{"rendered":"

Launched by Alibaba on August 3, 2026, Qwen3.8-Max combines a Mixture-of-Experts architecture with 2,400 billion parameters—95 billion of which are activated—with a reported context window of one million tokens and capabilities…

\n","protected":false},"author":7,"featured_media":658524,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"content-type":"","footnotes":""},"categories":[63,43],"tags":[59],"class_list":["post-658518","post","type-post","status-publish","format-standard","has-post-thumbnail","category-ia-agentique","category-innovation-competitivite-par-lia","tag-parlonsia"],"yoast_head":"\nQwen3.8-Max: How Alibaba's Model Is Really Changing Coding and AI Agents - aivancity blog\n\n\n\n\n\n\n\n\n\n\n\n\n\t\n\t\n\t\n\n\n\n\t\n\t\n\t\n\n","yoast_head_json":{"title":"Qwen3.8-Max : ce que le modèle d’Alibaba change réellement pour le codage et les agents IA - aivancity blog","description":"Qwen3.8-Max combines MoE architecture, a context of one million tokens, and AI agents. Analysis of performance, limitations, costs, and ethical considerations.","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https://aivancity.ai/blog/qwen3-8-max-codage-agents-ia-performances-limites/","og_locale":"fr_FR","og_type":"article","og_title":"Qwen3.8-Max : ce que le modèle d’Alibaba change réellement pour le codage et les agents IA - aivancity blog","og_description":"Qwen3.8-Max combine architecture MoE, contexte d’un million de tokens et agents IA. Analyse des performances, limites, coûts et enjeux responsables.","og_url":"https://aivancity.ai/blog/qwen3-8-max-codage-agents-ia-performances-limites/","og_site_name":"aivancity blog","article_published_time":"2026-09-22T09:56:09+00:00","article_modified_time":"2026-09-22T13:59:32+00:00","og_image":[{"width":1024,"height":1024,"url":"https://aivancity.ai/en/blog/wp-content/uploads/2026/09/lKimi-K3Architecte-Cloud2.png","type":"image/png"}],"author":"aivancity","twitter_card":"summary_large_image","twitter_misc":{"Écrit par":"aivancity","Durée de lecture estimée":"12 minutes"},"schema":{"@context":"https://schema.org","@graph":[{"@type":"Article","@id":"https://aivancity.ai/blog/qwen3-8-max-codage-agents-ia-performances-limites/#article","isPartOf":{"@id":"https://aivancity.ai/blog/qwen3-8-max-codage-agents-ia-performances-limites/"},"author":{"name":"aivancity","@id":"https://aivancity.ai/blog/#/schema/person/70f8508e84e45571c5fd172ea40ef3d4"},"headline":"Qwen3.8-Max : ce que le modèle d’Alibaba change réellement pour le codage et les agents IA","datePublished":"2026-09-22T09:56:09+00:00","dateModified":"2026-09-22T13:59:32+00:00","mainEntityOfPage":{"@id":"https://aivancity.ai/blog/qwen3-8-max-codage-agents-ia-performances-limites/"},"wordCount":2448,"image":{"@id":"https://aivancity.ai/blog/qwen3-8-max-codage-agents-ia-performances-limites/#primaryimage"},"thumbnailUrl":"https://aivancity.ai/blog/wp-content/uploads/2026/09/lKimi-K3Architecte-Cloud2.png","keywords":["Parlons IA"],"articleSection":["IA Agentique","Innovation & compétitivité par l’IA"],"inLanguage":"fr-FR"},{"@type":"WebPage","@id":"https://aivancity.ai/blog/qwen3-8-max-codage-agents-ia-performances-limites/","url":"https://aivancity.ai/en/blog/qwen3-8-max-codage-agents-ia-performances-limites/","name":"Qwen3.8-Max : ce que le modèle d’Alibaba change réellement pour le codage et les agents IA - aivancity blog","isPartOf":{"@id":"https://aivancity.ai/blog/#website"},"primaryImageOfPage":{"@id":"https://aivancity.ai/blog/qwen3-8-max-codage-agents-ia-performances-limites/#primaryimage"},"image":{"@id":"https://aivancity.ai/blog/qwen3-8-max-codage-agents-ia-performances-limites/#primaryimage"},"thumbnailUrl":"https://aivancity.ai/blog/wp-content/uploads/2026/09/lKimi-K3Architecte-Cloud2.png","datePublished":"2026-09-22T09:56:09+00:00","dateModified":"2026-09-22T13:59:32+00:00","author":{"@id":"https://aivancity.ai/blog/#/schema/person/70f8508e84e45571c5fd172ea40ef3d4"},"description":"Qwen3.8-Max combines MoE architecture, a context of one million tokens, and AI agents. Analysis of performance, limitations, costs, and ethical considerations.","breadcrumb":{"@id":"https://aivancity.ai/blog/qwen3-8-max-codage-agents-ia-performances-limites/#breadcrumb"},"inLanguage":"fr-FR","potentialAction":[{"@type":"ReadAction","target":["https://aivancity.ai/blog/qwen3-8-max-codage-agents-ia-performances-limites/"]}]},{"@type":"ImageObject","inLanguage":"fr-FR","@id":"https://aivancity.ai/blog/qwen3-8-max-codage-agents-ia-performances-limites/#primaryimage","url":"https://aivancity.ai/en/blog/wp-content/uploads/2026/09/lKimi-K3Architecte-Cloud2.png","contentUrl":"https://aivancity.ai/blog/wp-content/uploads/2026/09/lKimi-K3Architecte-Cloud2.png","width":1024,"height":1024},{"@type":"BreadcrumbList","@id":"https://aivancity.ai/blog/qwen3-8-max-codage-agents-ia-performances-limites/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Accueil","item":"https://aivancity.ai/blog/"},{"@type":"ListItem","position":2,"name":"Qwen3.8-Max : ce que le modèle d’Alibaba change réellement pour le codage et les agents IA"}]},{"@type":"WebSite","@id":"https://aivancity.ai/blog/#website","url":"https://aivancity.ai/en/blog/","name":"aivancity blog","description":"Advancing Education in Artificial Intelligence","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://aivancity.ai/blog/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"fr-FR"},{"@type":"Person","@id":"https://aivancity.ai/blog/#/schema/person/70f8508e84e45571c5fd172ea40ef3d4","name":"aivancity","image":{"@type":"ImageObject","inLanguage":"fr-FR","@id":"https://secure.gravatar.com/avatar/0b60f844cf48367ece3a9988562f25406b914c56b83ccd3df68e4c07737dc27e?s=96&d=mm&r=g","url":"https://secure.gravatar.com/avatar/0b60f844cf48367ece3a9988562f25406b914c56b83ccd3df68e4c07737dc27e?s=96&d=mm&r=g","contentUrl":"https://secure.gravatar.com/avatar/0b60f844cf48367ece3a9988562f25406b914c56b83ccd3df68e4c07737dc27e?s=96&d=mm&r=g","caption":"aivancity"},"url":"https://aivancity.ai/en/blog/author/bouazizaivancity-ai/"}]}},"amp_enabled":true,"_links":{"self":[{"href":"https://aivancity.ai/blog/wp-json/wp/v2/posts/658518","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https://aivancity.ai/blog/wp-json/wp/v2/posts"}],"about":[{"href":"https://aivancity.ai/blog/wp-json/wp/v2/types/post"}],"author":[{"embeddable":true,"href":"https://aivancity.ai/blog/wp-json/wp/v2/users/7"}],"replies":[{"embeddable":true,"href":"https://aivancity.ai/blog/wp-json/wp/v2/comments?post=658518"}],"version-history":[{"count":8,"href":"https://aivancity.ai/blog/wp-json/wp/v2/posts/658518/revisions"}],"predecessor-version":[{"id":658538,"href":"https://aivancity.ai/blog/wp-json/wp/v2/posts/658518/revisions/658538"}],"wp:featuredmedia":[{"embeddable":true,"href":"https://aivancity.ai/blog/wp-json/wp/v2/media/658524"}],"wp:attachment":[{"href":"https://aivancity.ai/blog/wp-json/wp/v2/media?parent=658518"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https://aivancity.ai/blog/wp-json/wp/v2/categories?post=658518"},{"taxonomy":"post_tag","embeddable":true,"href":"https://aivancity.ai/blog/wp-json/wp/v2/tags?post=658518"}],"curies":[{"name":"wp","href":"https://api.w.org/{rel}","templated":true}]}}