Tencent launched Buddy One at its 2026 Global Digital Ecosystem Conference, connecting WorkBuddy with Tencent Docs, Tencent Meeting, Tencent Drive, and ima through shared accounts, authorized cross-app search, storage, and AI points. WorkBuddy can search authorized documents, meeting records, and knowledge bases to produce answers labeled with their sources, while Tencent Drive opened a beta application as shared storage for AI-generated assets. Tencent also plans to open WorkBuddy's architecture to enterprise-developed and third-party Agents, but broader cross-app memory sharing remains under development.
Tencent WorkBuddy has launched a feature that lets users connect to a computer from a phone, view the desktop in real time and take control of the computer. Users can also see task progress and results on the phone.
Youtu-Parsing-Omni is a 5B-parameter parser designed to convert documents, images, audio, and video into structured records grounded in space and time. It combines a shared-weight encoder, cross-modal fusion, a unified output schema, and self-evolving post-training across ten parsing tasks. The authors report a 96.96 overall score on OmniDocBench v1.6 and a 75.08 average on OmniParsingBench, while code and model weights are publicly available.
A Z Finance report says ByteDance created a company-level AI Data and Security department overseeing cross-modal data sourcing, procurement, synthesis, cleaning, quality evaluation and delivery. ByteDance Seed separately established a Pretrain Data department led by Li Chenggang. Tencent also changed its AI Data leadership, while Xiaohongshu hired former early DeepSeek member Chen Guanting to lead pretraining data. Separately, AI evaluation and synthetic-data startup UniPat AI was reportedly about to complete roughly $300 million in financing at a $2.5 billion valuation, with Alibaba proposed to lead and Tencent and Sequoia China among the prospective investors. The financing had not been confirmed.
The Information reported that DeepSeek raised 105.5 billion yuan in a second funding round at a valuation of 615.5 billion yuan, with nearly 30 investors including Tencent, CATL, ChangXin Storage, SMIC, Meituan, vivo and Xiaohongshu. State-backed investors reportedly contributed about 27%, up from 2% in the first round. The Information attributed the expanded fundraising target to rising Nvidia chip prices and preparations for a possible AI-industry downturn. A separate 36Kr report said the financing closing was still awaiting official disclosure.
Beijing’s market regulator organized Baidu, Ocean Engine, Tencent, Kuaishou, JD.com, Meituan, 360 and ifeng.com to sign the Internet Marketing and Promotion Self-Regulation Convention. The industry-led pact covers commercial advertising and promotion through text-and-image posts, short videos and livestreams, and sets four principles: value orientation, authenticity, fair competition and safety compliance. Platforms must verify advertiser identities, qualifications and administrative permits, retain electronic data for at least three years, and conduct pre-publication, in-process and post-publication reviews. Livestream platforms must review participants and monitor livestreams. Commercial content made with AI deep synthesis must be prominently labelled, and platforms may not use AI to create false information, false endorsements or violations of others’ rights. The pact also addresses misleading promotions, consumer complaints, minor protection and rules for healthcare, education, finance, pharmaceuticals, medical devices, health foods and real estate. Signatories said they would assist regulatory investigations and provide relevant data.
QuestMobile data cited by TMTPost put WorkBuddy’s September multi-platform monthly active users at 32.35 million, ranking first among office agents, while Enterprise WeChat and WorkBuddy completed account integration on October 14. The same reports highlighted Green Bubble’s WeChat mini-program reach of 89.68 million weekly users in early October and framed Tencent’s strategy as adding AI and entertainment within existing workflows rather than launching standalone apps. These figures come from QuestMobile and company disclosures reported in analytical columns rather than a primary product announcement.
TMTPost interviewed five former heavy gamers who now spend more time watching AI short dramas, citing easier access, lower commitment, stronger emotional payoff and fewer interruptions than games. The report says Red Fruit's core users overlap heavily with gamers and average daily use exceeds two hours, but provides no methodology or independent verification for that claim. Predictions that AI short dramas will reduce gaming engagement or spending remain unverified analyses based on a small set of personal accounts.
cnBeta and TechNews reported that Alibaba, Tencent, and ByteDance are seeking long-term supply agreements with Samsung Electronics and SK hynix as AI data-center demand increases. The reported agreements could run for up to 10 years and focus on server DDR, LPDDR, and other DRAM or NAND supply, but neither outlet confirmed that any agreement has been signed. TechNews cited industry sources and an executive who said the Korean suppliers retain advantages in high-performance memory and mass-production stability, while cnBeta also cited reports that CXMT and YMTC cannot yet fully meet cloud providers’ requirements.
DoNews reports that game companies that entered AI short dramas and comic dramas are now testing their ambitions against performance data as short-term KPIs and uncertain returns pressure the projects. The report says most entrants lack rigorous attribution, operation, and settlement standards, although isolated examples include billion-level views and campaigns that successfully fed users back into games. An authorized reprint by 创业邦 carries the same analysis rather than providing independent corroboration.
IT之家, citing Maeil Business Newspaper, reports that Alibaba, Tencent, ByteDance and other Chinese cloud providers are negotiating long-term server-memory supply agreements with Samsung Electronics and SK hynix. The talks reportedly cover server-grade DDR and LPDDR, with some DRAM customers seeking ten-year terms, although no finalized agreement has been confirmed. Wccftech characterized the arrangements as ten-year LTAs and linked the negotiations to tight server-memory supply.
Huxiu AI’s republished commentary examines a wave of researcher departures, highlighting Chen Guangyu’s announced exit from Kimi, Lin Junyang’s public separation from Qwen, and unconfirmed reports that Google DeepMind researcher Yao Shunyu is preparing to found a startup. It argues that rapid recognition has given AI researchers greater bargaining power while prompting companies to rely on non-compete clauses, delayed departures, restricted publicity, and richer compensation packages. The article frames these cases as a broader contest for scarce AI talent rather than isolated moves.
TMTPost reports that Tencent has embedded the Green Bubble short-drama service in WeChat and introduced incentives aimed at attracting creators. A hands-on comparison found overlapping titles with Red Fruit, weaker related-content suggestions, and social visibility around likes and favorites that may deter private viewing. The report also identifies AI video generation as an important driver of the short-drama boom.
Kling AI affiliate Beijing Kling Intelligent Technology Co., Ltd. added Shanghai Qishan Investment, a Tencent subsidiary, Hangzhou Alibaba Cloud Feitian Information Technology Co., Ltd., Beijing Baidu Netcom Science and Technology Co., Ltd., and National AI Industry Investment Fund Partnership (Limited Partnership) as shareholders, according to 36氪 citing the Tianyancha app. Its registered capital increased from approximately RMB 71.924 million to RMB 87.447 million, a rise of about 22%.
Analysysys’ September 2026 China desktop AI-native office-agent ranking, relayed by 36Kr, put Tencent’s WorkBuddy first with 85.34 million monthly visits. Alibaba’s Qwen Office followed with 46.55 million visits, ahead of ByteDance’s Doubao Work at 30.39 million, TRAE at 27.12 million and CodeBuddy at 25.88 million; Baidu’s Dazi was reported eighth with 9.26 million interactions.
Reuters reports that DeepSeek's latest funding round has reached about RMB100 billion, roughly $15 billion, at a pre-money valuation of RMB500 billion. Tencent and CATL are among the largest investors, while Meituan, Geely and GigaDevice are also participating, according to Reuters sources. Completion is expected in the coming days, but the terms could still change.
An analysis from the Bone Editorial team at the Chinese WeChat account “骨朵” says live-action short-drama categories expanded from 44 to 65 within 10 months in 2025. In 2026, AI short dramas added themes such as high-concept premises, food and restaurant businesses, stockpiling, science-fiction mecha, and social issues. The analysis argues that AI short dramas use large-scale content experimentation to test whether a theme can work in weeks or days, while long-form productions continue to operate on a multiyear timeline for planning, filming, and release. It examines no-CP female-led stories, system-based farming, high-concept premises, cyberpunk, and infinite-flow as subjects short dramas can test while long-form projects consider whether and how to pursue them.
Red Fruit Short Drama announced a phased withdrawal of its coin system, stopping coin issuance for new users from November 1 and ending the feature for all users before year-end; it also added rest and sleep reminders on October 23. Tencent’s Green Bubble Short Drama is competing for creators through a campaign offering eligible new dramas up to 100% revenue sharing and by adapting WeChat public-account novels into micro-dramas. TMTPost interprets the contrasting approaches as a shift from rewarding viewing time toward competing through content quality, while noting that AI-generated short dramas face thin profitability and uncertain platform incentives.
Tencent WorkBuddy announced that it has integrated Kimi-K3.1-Pro and Kimi K3.1 Pro Highspeed, according to IT之家 and 36氪快讯. The announcement says the standard model supports a 1 million-token context window, while the Highspeed variant offers equivalent intelligence and approximately two to three times the output speed. Because resources are tight, access is currently prioritized for Enterprise Flagship and personal subscription users.
Tencent released the 9B Hy-Nomas-Diffusers package on Hugging Face under the Apache-2.0 license. It provides BF16 Diffusers weights and the HyNomasModularPipeline. According to the model card, Hy-Nomas supports multi-turn text interaction, camera control and continuous generation through a persistent KV cache. Tencent also released Hy-Nomas-Bench, a four-part benchmark suite with 280 test cases covering text interaction, camera navigation, combined text and camera control, and spatial memory.
TMTPost reports that Meta’s Q3 2026 10-Q disclosed up to about $824.9 billion in operating and finance leases, irrevocable purchase contracts, and cloud-capacity commitments extending through 2036. The figure combines distinct obligations and was not presented by Meta as a single $825 billion AI bill. Meta also reported $32.79 billion in operating cash flow against $34.91 billion in capital expenditures, leaving quarterly free cash flow at negative $2.12 billion.
A report from 硅谷101 examines a new wave of AI personal assistants, including Anthropic's Claude Duet, Meta's Muse, SpaceX's Grok Bot, OpenAI's Dots and Instinct. The products use different entry points and execution environments to handle multi-step tasks, while the analysis highlights unresolved questions around costs, privacy, platform access and revenue. It reports that Instinct, Muse and Grok Bot had not disclosed clear revenue or profit figures, and that the market remains at an early stage.
36氪 reported that Alibaba, ByteDance, JD.com, Tencent, Kuaishou, and Tmall advanced AI shopping and merchant tools during the 2026 Singles’ Day shopping festival. Named actions included the release of Qwen Office’s e-commerce edition, Doubao’s dedicated Singles’ Day entry, an expanded JD Jingmai AI Business Center, AI-assisted selection for WeChat Shops, and AI operating tools for Kuaishou merchants. The report said these products extend AI from background copy and customer-service assistance into shopping guidance, business management, and purchasing decisions.
Reports from TMTPost and 36Kr, both syndicating an analysis by AI大模型工场, describe a rapid expansion of China’s personal-agent market. Alibaba is advancing Qwen Go, Tencent is exploring a standalone personal agent, ByteDance is extending its phone-system agent efforts, StepFun has launched an agent phone, Ant Group is upgrading its assistant, and Today AI disclosed more than $200 million in financing. The analysis says ByteDance’s Doubao Phone Assistant listed 155 applications refusing automated operation, including WeChat, Taobao, and Meituan, highlighting tension between companies seeking to become new user entry points and platforms restricting access to existing services.
According to a First Financial report republished by 36Kr, southbound funds recorded net selling of HK$1.670 billion on October 23, 2026. Zhipu ranked first in net buying with HK$1.739 billion, followed by Tencent at HK$1.296 billion and 南方恒生科技 at HK$580 million. Alibaba-W, CNOOC and SMIC recorded net selling of HK$1.406 billion, HK$801 million and HK$799 million, respectively.
Tencent’s WeChat Vision released WeVisOne-VL-2B, a 2B-parameter image-text-to-text model built on Qwen3-VL and licensed under Apache 2.0. According to the model card, it combines object-level tasks such as open-vocabulary detection, referring localization, counting, pointing and OCR with pixel-level referring segmentation and promptable alpha matting. The card also warns that fine alpha boundaries can be sensitive to image resolution, occlusion, transparency and prompt phrasing.
Tencent WeChat Vision released WeVisOne-VL-4B, a Qwen3-VL-based vision-language model designed to combine object-level perception with pixel-level tasks. Tencent’s model card lists open-vocabulary object detection, referring localization, object counting, object pointing, OCR localization, referring segmentation, and promptable alpha matting. The release includes an inference package and a required 9.6-million-parameter pixel decoder. Tencent says fine alpha boundaries are sensitive to image resolution, severe occlusion, complex transparency, and prompt phrasing, and that the model currently targets single-turn image perception. Unless otherwise noted, Tencent-authored code is released under Apache License 2.0; third-party components retain their original licenses and copyright notices.
Tencent announced that QClaw will stop operating on December 24, 2026, after reaching public beta on March 20. The company will provide a tool for migrating users’ data to WorkBuddy, while Qwen Go! is an announced personal assistant intended to handle tasks such as flight-delay claims, price comparisons and phone bookings; the latter capabilities were not independently tested.
The Wall Street Journal, citing people familiar with the matter, reported that DeepSeek completed a second private funding round raising US$15 billion, about RMB100.323 billion. Strong investor demand reportedly expanded the round to nearly twice its initial target. Before the round, DeepSeek was valued at about US$74 billion. CATL and Tencent reportedly invested again, while new participants included semiconductor-focused Shixi Capital, Lenovo-backed Legend Capital and automaker Geely. Unnamed sources also said DeepSeek is accelerating preparations for an IPO in Shanghai early next year.
TMTPost reports that multimodal models such as Seedance and MiniMax H3 are being integrated into content-production workflows for advertising, gaming, e-commerce and interactive media. The report says growing demand for channel-specific and frequently refreshed assets is outpacing conventional production, making AI adoption increasingly important for repetitive, high-volume tasks. It describes an unnamed large game company using a self-built technology platform and MiniMax H3 and Seedance to produce marketing images and videos, reducing some production times and costs. MiniMax H3 was also used in Bilibili’s updream creator platform, while other platforms such as Workbuddy and Miora were adjusting their tools for new models. The report argues that practical adoption depends on reliability, controllable costs, creative consistency, lower review and rework rates, and campaign feedback—not technical benchmarks alone. It also describes interactive film-game formats as an emerging direction whose sustainable commercial cycle has not yet been established.
What problem it solves
AI agents need isolated environments that can execute code and workloads while supporting fast creation, secure network access, and manageable state. They also need an interface that can integrate with existing agent tooling and deployment workflows.
Key features
Hardware-level isolation — Each sandbox runs in its own MicroVM with a dedicated operating system kernel. Kernel-…
What problem it solves
Real-world translation involves more than converting text between languages: tasks can require terminology, style, personalization, delimiter preservation, and structured-data constraints. Hy-MT2 is designed to support translation among 33 languages and follow translation instructions in multiple languages for complex real-world scenarios.
Key features
Instruction-driven tas…
What problem it solves
A coding-agent run can skip planning, omit tests, review its own work, or stop before completion, making delivery inconsistent and difficult to supervise. Teams need repeatable workflows that separate roles, execute work in sandboxes, preserve review evidence, and require human approval before changes ship.
Key features
Repeatable SOP delivery — Standard operating procedures…
What problem it solves
Teams need to bring documents together so they can be searched, reasoned over and kept up to date. They also need a way to move from source-grounded retrieval to multi-step tasks while maintaining organized, current knowledge.
Key features
Three knowledge workflows — RAG supports lookup, an agent handles multi-step tasks, and a wiki organizes knowledge. All three operate on …