Today we’re releasing GDPx, a benchmark for measuring agentic capabilities across the most valuable domains. GDPx is a true frontier-level benchmark: even the best model scores just 12% (1/7)🧵 https://t.co/RNnkcuUkih
#English text is being prepared.#Evaluations / benchmarks
English text is being prepared.
#Industry news#Agents#English text is being prepared.
English text is being prepared.
#Product updates#Video#Multimodal
Schneider Electric has released the RapidScale modular data center reference design, purpose-built to adapt to the NVIDIA Vera Rubin platform, incorporating liquid cooling technology and integrating power, cooling, IT, and full-lifecycle services. Real-world testing with Compass Datacenters shows that prefabricated IT modules can reduce on-site installation time per module from 75.5 hours to 15 hours, a reduction of 80%. The design has been validated according to ANSI standards and is applicable to the U.S. market, with plans to expand to IEC international standards in the future.
#Product updates#English text is being prepared.#English text is being prepared.
According to QUICK FactSet data, in the third quarter of 2026, the five major memory chip manufacturers—Samsung Electronics, SK Hynix, Micron Technology, Kioxia, and SanDisk—had a combined operating profit of approximately 32 trillion yen, a 10-fold increase compared with the same period last year, surpassing the combined approximately 25 trillion yen of Alphabet, Microsoft, Amazon, Meta, and Apple. Samsung Electronics' operating profit for the quarter was 107.44 trillion won, up about 783% year on year; SK Hynix's operating profit was 83.33 trillion won, with an operating margin of 80%. The profit growth mainly came from demand for HBM from AI data centers; manufacturers shifting capacity to HBM squeezed the supply of general-purpose DRAM, and Samsung's average DRAM selling price in the third quarter rose about 20% quarter on quarter, while HBM4 sales grew more than threefold quarter on quarter.
#Industry news#English text is being prepared.
Zhuoyu Technology released its commercial-vehicle intelligent assisted driving solution, Zhuoyu Pansing, in Zhuhai. It is divided into two versions, PDS 4 and PDS 5. PDS 4 supports highway NOA and has already entered mass production across multiple brands, while PDS 5 corresponds to a native multimodal foundation model and will support urban NOA. Pansing has partnered with six commercial heavy-truck brands—XCMG Automobile, Shaanxi Automobile Group, Sinotruk, FAW Jiefang, Foton Motor, and Dongfeng Commercial Vehicle—which together account for more than 93% of China's commercial heavy-truck market. Zhuoyu says its AI reinforcement-learning fuel-saving algorithm achieved a measured fuel-saving rate of 4.5%–7.5% in tracking tests covering more than 200 mass-produced vehicles and cumulative mileage of over 20 million kilometers.
#Product updates#English text is being prepared.#English text is being prepared.
The future training center results released by the Ministry of Education in October 2026 list "sorting out and stabilizing the competency map of key on-site positions" as an important cornerstone. The map determines what AIGC training labs teach and how they assess; if it is made into a static table, the curriculum degenerates into software operation training. The implementation path is to pedagogize production tasks, ladder position competencies, and curricularize industry standards. On the assessment side, automatically collectable process evidence must be retained for each competency dimension. The Edu.AI platform of Shikong Jiepai (Hangzhou) carries subject workflows and teaching management according to this approach. A 50-workstation plan dated October 2026 totals 1,389,600 yuan including tax, or 27,792 yuan per student.
#Industry news#English text is being prepared.
The three major A-share indices rose collectively. The ChiNext Index gained 3.28% to 2,492.91 points, the Shanghai Composite Index rose 1.16%, and the STAR 50 rose 2.95%. A total of 4,783 stocks in the whole market advanced, with 93 hitting the daily limit-up, and half-day turnover across the two markets reached 1.1401 trillion yuan. On the board, AI video, gaming, Zhipu AI concept, Kimi concept, media, and film and cinema chains led the gains, while precious metals, coal mining, banking, and oilfield engineering services fell against the trend.
#Industry news#Video
Du Xiaoman, the Hong Kong University of Science and Technology, and the Hong Kong Generative AI Research and Development Center (HKGAI) jointly released the Xuanyuan 4.5 financial-enhanced large model at the opening ceremony of the 2026 Hong Kong FinTech Week. Based on RSI technology, it improves effectiveness by 30% in core financial scenarios such as customer acquisition, risk control, and customer management, while costing only 10% of a general-purpose model. The model consists of two parts: a large language model and a multimodal model. The self-developed Xuanyuan-Machine participates in agent data synthesis, supervised fine-tuning, reinforcement learning, and other stages, and has built more than 100 billion tokens of scenario-enhanced data. Based on Xuanyuan, Du Xiaoman has built over a thousand agents and more than 5,000 skills, with average daily token usage exceeding 100 billion, and has shortened the iteration cycle of risk strategies in its credit business by more than half.
#Model releases#Agents#English text is being prepared.
Xuhui embodied intelligence company Daxiao Robotics, in collaboration with Nanyang Technological University S-Lab, released the world model Mundo, which requires only 1 to 12 real photographs or short videos to reconstruct indoor 3D spaces such as offices and home rooms, and supports free roaming and physical collision interaction. The model can write physical parameters such as object mass and friction into the scene; users can pull open drawers, push tables and chairs, and throw objects through code, and the physics engine calculates motion and collision results. The system can also synchronously output the robot's first-person and third-person views and retain motion trajectories and interaction data. Mundo is a supplement to Daxiao Robotics' "Kairos" world model system. The company has attempted to reconstruct a hotel floor environment, simulating tasks such as opening doors and picking up and placing objects to carry out robot simulation training. The relevant project page and technical blog have been launched.
#Model releases#Embodied AI#Multimodal
Meituan has released its next-generation large model LongCat-2.5, a natively multimodal model designed for long-horizon agent tasks, with 1.6 trillion total parameters, trained entirely on domestic compute. During the mid-training stage, the model conducted long-context training with sequences of up to 1M tokens, with cumulative training volume reaching trillions of tokens, and it optimized task division around supernodes to address the challenge of small per-card memory in domestic compute. Meituan says its results on multiple mainstream benchmarks fall between Claude-Opus-4.8 and Claude-Opus-5, and it surpasses Claude-Opus-5 on some evaluations.
#Model releases#Multimodal#Agents
Xu Guanju, Chairman of Transfar Group, proposed at Transfar's 40th Anniversary Global Partners Conference that over the next decade Transfar Group aims to leap toward becoming a world-class enterprise, transforming with an AI-native approach that runs through the entire corporate governance and management process. Xu Guanju said that over the past decade, Transfar's ranking among China's Top 500 Enterprises rose from 426th to 191st, with total operating revenue growing fivefold and net assets growing threefold. Transfar's various business segments will advance AI-driven design and closed-loop experimentation on the R&D side, AI optimization on the production side, global intelligent scheduling of the supply chain, and agent-based services on the platform side.
#Industry news#Agents
According to foreign media citing people familiar with the matter, Meta has collaborated with Chinese robotics company Fourier to develop a humanoid robot prototype named Max, and the two sides are developing a second-generation Max 2, with plans to redesign the torso and hands and add sensors to assist with going up and down stairs. Fourier told The Paper on November 2 that if there is any latest news later, it will be announced through official channels; a Meta spokesperson said the collaboration focuses on internal research and experimentation, and there are currently no plans to sell the prototype. Meta has also recently been recruiting for robot hardware R&D and supply chain positions in Shanghai and Shenzhen, and in May this year acquired humanoid robot foundation model company Assured Robot Intelligence.
#Industry news#Embodied AI#Meta
Weekend hack: HackerTube. I was annoyed with YouTube's recommendations so I had Claude filter down 43M comments over 20 years of HN to 6,454 videos that keep being recommended. *Significantly* better viewing experience, ~2 hours with Opus 5.5: https://t.co/JIhXN7OSh1 - Excellent recs without any logins (state is kept in local browser session) - Thumbnail is a representative frame from middle of the video instead of the super duper A/B test hyper-optimized user uploads - Underlying data from the comment analysis as one zip file, point your Claude at it and ask for your own version - HackerTube Longs ✊ A bit more on the process: I use Claude Code (desktop app for Mac), Opus 5.5 on Medium effort as my default. I only talk to Claude and have it do everything, the only "manual work" I helped with in tandem was in the browser (reminiscent of MenuGen): 1) I had to create some kind of API key from Google Cloud UI to query YouTube for video metadata, 2) I booted up a VPS box for serving and 3) I bought the domain name. Once the box was up I gave Claude the IP and told it update, secure, harden, deploy and optimize the code, and it did the whole thing. Many awesome projects that would have normally taken weeks/months even can now be few-shotted for breakfast. What else are we not thinking of.
#English text is being prepared.#Agents#English text is being prepared.
#English text is being prepared.#Image generation
The 2026 Alibaba Cloud Double 11 promotion covers two product lines, cloud servers and AI large models, forming a multi-tiered structure of cloud server special offers, AI large model discounts, Token Plan subscription discounts, coupons, free trials, and bundle purchases. The lightweight application server with 2 cores and 2 GB is available at a flash-sale price of 38 yuan/year, the special-offer ECS Economy e instance at 99 yuan/year, and the General-Purpose Computing u1 instance at 199 yuan/year. Calls to Qwen3.8-Max and Qwen3.8-Flash from 22:00 each night to 8:00 the next morning enjoy a 40% discount, and AI experience coupons are valid for 14 days after being claimed.
#Industry news#OpenAI
The goal of Generative Engine Optimization (GEO) is not link rankings, but making content the one sentence that AI extracts as the answer. The article breaks down the six gates AI passes through when answering questions—intent rewriting, retrieval recall, snippet ranking, cross-verification, answer generation, and source citation—and distills six trust variables: extractability, entity clarity, multi-source consistency, timeliness, professional depth, and semantic coverage. The author emphasizes that GEO can only increase the probability of being cited, not guarantee it, and that brand-term pollution and low-quality content flooding must also be guarded against.
#Tutorials / practice#Search#RAG
The AI portrait fashion editorial generation service on the Alibaba Cloud Marketplace API Gateway is integrated as an asynchronous task: authenticate with APPCODE, submit img_url, theme, and aspect_ratio to immediately receive a task identifier, then poll for the result. A single submission can produce multiple sets of styled editorial shots based on the same portrait, each accompanied by an English prompt, with aspect ratios supporting 1:1 / 3:4 / 4:3 / 16:9 / 9:16. When theme is left blank, the model automatically plans a theme based on the subject's temperament.
#Tutorials / practice#Image generation#English text is being prepared.
Home-school mini programs can integrate an AI fitness testing plugin that uses camera frame sampling, recognition of 17 human body keypoints, and motion analysis to achieve automatic counting, posture assessment, and scoring correction for more than ten types of exercises such as jump rope and sit-ups. The entire pipeline completes within a few hundred milliseconds on the phone, without uploading videos or relying on GPU servers, and can be integrated and launched in as fast as one to three days. The solution covers daily check-ins, automatic grading of fitness test homework, and movement guidance, and results can be exported with one click for physical health test data reporting.
#Tutorials / practice#English text is being prepared.#Multimodal
Alibaba Cloud's 2026 Double 11 promotion features a Lightweight Application Server with 2 vCPUs, 2 GB of memory, peak 200M bandwidth, and a 40G ESSD cloud disk at a flash-sale price of 38 yuan/year as the lowest threshold, along with 2 vCPUs/4 GB at 9.9 yuan/1 month and 199 yuan/1 year. General-purpose compute u2a instances start at 701.40 yuan/year for 2 vCPUs/8 GB (1M bandwidth) up to 1,007.40 yuan (5M bandwidth), while 4 vCPUs/16 GB ranges from 1,291.80 yuan to 1,597.80 yuan. The 9th-generation enterprise ECS c9i/g9i/r9i are equipped with Intel Xeon 6 processors, delivering up to a 20% maximum improvement in single-core computing power compared with the 8th generation.
#Industry news