qianwen

Alibaba's Qianwen Office product has been online for a month, with the number of users exceeding 30 million, more than half of whom are enterprise users. In the past 30 days, a total of 120 version updates have been completed, averaging 4 iterations per day, with thousands of features updated cumulatively. The international version has also been launched.Qianwen Office's open-source enterprise context infrastructure MyContext has received over 3,000 stars on GitHub within a few weeks of being open-sourced. Combined with deep integration with IM software like DingTalk, it can transform information scattered across business processes such as messages, documents, approvals, emails, and knowledge bases into organizational context that is understandable and callable by agents. At the same time, the Qianwen large model team has launched a dedicated office version Qwen3.8-Flash, specifically optimized for multi-step planning, tool selection, and context compression, achieving approximately a 100% increase in single-task generation speed in real office scenarios, with average token consumption reduced by 75%.Companies such as Changan Automobile, Huifu Tianxia, Chuanhua Group, Zhongyi Co., and Laoxiangji have already integrated. At Changan Automobile, Qianwen Office has streamlined the entire supply chain from research and development to sales and service, reducing the manual wiring harness selection calculation by senior engineers from two days to 5 minutes, and the verification of vehicle documents from over ten minutes to 1-2 minutes.

The IDC report "China Retail Cloud Market Share, 2025" shows that Alibaba Cloud ranks first with a market share of 30.4%. Infrastructure is the largest component of the retail cloud market, with Alibaba Cloud maintaining the top position with a 37.5% share.Alibaba Cloud provides a comprehensive AI technology system covering "Chip---Cloud---Model---Application" to retail enterprises, offering products and services that include infrastructure, data platforms, large models, and industry applications. The integration of AI technology with retail business processes covers scenarios such as smart marketing, intelligent shopping guides, smart customer service, user insights, product operations, and supply chain management.Adidas has created a dedicated AI product design platform based on Alibaba Cloud's artificial intelligence platform PAI and the Qianwen large model. Anta, in collaboration with Alibaba Cloud, has developed the "AI Planning" intelligent agent platform based on the Lingyang AgentOne and the Qianwen large model. Coach, L'Oréal, LAWSON, China Duty Free Group, and Xiaopei Pets have also accessed related capabilities for intelligent upgrades.

Ali Qianwen released the Qwen3.8-Flash model, which is described as a multimodal MoE model and an early preview of the Qwen4 architecture.The production version of Qwen3.8-Flash will soon be available through the Qwen Cloud API, with a price of only $0.16 per million input tokens and $0.47 per million output tokens. The model has 125 billion parameters + 51 billion N-gram embedding parameters, but only activates 6 billion parameters per token, achieving extremely high cost-effectiveness.

According to the "Science and Technology Innovation Board Daily," Alibaba Qianwen is currently testing a multi-Agent collaboration feature on the Qianwen creation platform in a small scope, mainly aimed at professional creative scenarios such as AI short dramas, comic dramas, marketing advertisements, and image design. Currently, only a small number of professional creators have been granted testing qualifications.This feature enables multiple Agents to collaborate, linking together the stages of creative planning, script writing, visual design, storyboard generation, and final production. Different Agents can cooperate based on task division to complete the entire production process from concept to final product. Users can participate in discussions throughout the process, adjust the creative direction at any time, or delegate some tasks to AI for independent completion.Recently, Qianwen has launched Qianwen creation for the professional content production field, and it has already integrated Alibaba's latest models such as Wan 3.0, Wan 3.0 Prime full version, and Qwen-Image 3.0 Pro.

As of August 24, Shanghai has added 1 generative artificial intelligence service that has completed filing, with the model name being Qianwen Task Assistant; a total of 188 generative artificial intelligence services have completed filing.

Qianwen's former technical director Lin Junyang announced the establishment of a new company, Pragmatic Technology, in Shanghai, focusing on the research direction of next-generation intelligent agents that span the digital and physical worlds.He also revealed that Gao Rong Venture Capital and HSG (Sequoia China) jointly led this round of financing, with support from Tencent and the Shanghai Future Industry Fund. According to informed sources, the "this round of financing" mentioned by Lin Junyang refers to the angel round financing of Pragmatic Technology, with a valuation of 2 billion dollars.

According to information from the official Apple website, Apple Smart has supported collaboration with Alibaba's Qianwen model.According to the official introduction, after users enable the relevant features, Siri can use the Qianwen model to provide more in-depth answers, including analyzing photos and documents; the writing tool can also leverage Qianwen to generate text or image content based on user descriptions.

According to exclusive reports from Huxiu, Alibaba has internally regarded "AI Office" as one of its strategic directions and has completed the integration of organization and products: the original Wukong Business Unit has been upgraded to "Qwen Office Business Unit," led by Chen Yusen, combining the three products QoderWork, Wukong, and MuleRun into Qwen Office (QwenWork), with the three products and teams no longer existing separately. Qwen Office began public testing on August 3, with the web version and PC client now available, and the DingTalk entry will be launched soon. On the same day, Alibaba released the Qwen3.8 model, enhancing Coding and Cowork capabilities.Alibaba positions Qwen Office as a "enterprise-level AI productivity platform aimed at corporate organizations," rather than a personal office tool, and will not focus on user numbers and monthly active users as core assessment targets in the short term, instead concentrating on product experience and enterprise scenario implementation. The Qwen Office Business Unit is also responsible for DingTalk, intending to leverage the million-level enterprise customer resources accumulated by DingTalk. According to data analysis from Analysys, in June, the desktop access volume of AI office intelligent agents was led by Tencent WorkBuddy with approximately 20.97 million visits, followed by Byte's TRAE IDE with about 12.79 million, and Alibaba's QoderWork with around 7.88 million (Wukong had an additional 1.31 million visits). This integration is viewed internally as a key step to catch up with Tencent and Byte. Following the news, Alibaba's Hong Kong stock rose by over 7% on August 3.

According to official news, Alibaba has released the Qianwen 3.8-MAX artificial intelligence model, which has 2.4 trillion parameters. The model will open source its weights next week, marking the first time Tongyi Qianwen has open-sourced a Max-level model, along with the simultaneous open-sourcing of Qwen3.8-27B.Qwen3.8-Max focuses on autonomous programming and multimodal capabilities, capable of over 10 days of self-evolution development and 500+ rounds of chip design optimization. The API pricing is $2.0/M tokens for input and $6.0/M tokens for output. This is the second open-source weight model with over 2T parameters following Kimi K3, marking an intensification of competition in the open-source large-scale model space.

According to Jinshi News, Qianwen Assistant has undergone extensive testing within Tesla's vehicle system and is expected to be integrated soon. Insiders reveal that Qianwen has completed a large number of tests in the real vehicle environment of Tesla, possessing the capabilities of "listening and responding, controlling the car, navigating, and handling tasks."In the future, the functions of Qianwen on Tesla's vehicle system may exceed the traditional definition of a "voice assistant." Currently, Alibaba Cloud and Tesla China have not made any official response to this.

According to a report by Reuters, the Cyberspace Administration of China announced on Wednesday (July 15) that Apple's edge-generative AI service "Apple Intelligence" has successfully completed filing in China and is approved for use on iPhone devices in the Chinese market. The official announcement has not yet specified a concrete launch date.According to informed sources, Apple's AI service will integrate the AI model capabilities of Baidu and Alibaba. Alibaba subsequently confirmed in a statement that its "Tongyi Qianwen" (Qwen) large model will be integrated into the Apple Intelligence experience for users in China on iOS, iPadOS, macOS, and visionOS. Apple has not yet responded to this.

Vitalik Buterin posted that there is a winner in the AI anonymous recognition challenge experiment he previously initiated. In this experiment, he wrote the Chinese version of EIP-7503 (Zero-Knowledge Wormhole) in 2024, manually corrected after using Qwen 2.5 for local translation, in an attempt to hide his identity.The article states that although the prose style was effectively blurred, the AI successfully identified it by detecting unique thinking habits in mathematics and algorithm explanations (such as specific numerical examples and logical chains), completely bypassing its obfuscation strategy that only targeted prose.

Ali Qianwen announces the open-source Qwen-Scope------an interpretability module trained based on the Qwen3 series and Qwen3.5 series models, with application scenarios including directed control of inference results, data classification and synthesis, model training and optimization, evaluation sample distribution analysis and comparison, etc.The weights of this open-source Qwen-Scope involve 7 large models, covering the dense models and mixture of experts models of the Qwen3 and Qwen3.5 series, with a total of 14 groups of sparse autoencoder weights.

The 0G Foundation announced a partnership with Alibaba Cloud to launch on-chain access to the Qianwen large model for AI agents.This move shifts artificial intelligence infrastructure from centralized, API-based systems to programmable, tokenized access specifically designed for autonomous agents. Through this collaboration, 0G allows developers to directly embed access to Qianwen into their infrastructure via a token-based mechanism. This eliminates the need for traditional account management and enables on-demand programmable access to AI computation. Agents running on 0G can query the Qianwen large model to generate responses, process information, or execute tasks in real-time.Michael Heinrich, CEO and co-founder of 0G Labs, stated, "From a macro perspective, inference runs on Qianwen, and verification runs on 0G—this builds a more complete computational and trust foundation for autonomous AI systems. The integration of 0G with Alibaba Cloud's Qianwen marks a milestone, achieving the fusion of top-tier intelligence and trusted infrastructure."

According to reports from the official public account, Alibaba's Qianwen App has completed an upgrade of its "in-depth research" feature, adding a financial analysis module. In collaboration with Tonghuashun, it integrates real-time minute-level market data for over 13,000 stocks and consolidates approximately 1 million financial reports, announcements, and authoritative research reports from listed companies. Users can generate a complete in-depth research report with one click using pre-set analysis instruction templates. The report supports automatic generation of visual charts and can be exported in Word/PDF format.The core technological support for this upgrade is the Qianwen Agentic architecture, which has autonomous planning and execution capabilities. It can interpret user intent, automatically call real-time data, and integrate multi-source information to output conclusions, with all key data clearly indicating the original source.