Tencent

Two months after the release of POINT1.0, Tencent announced the launch of POINTS1.5 on December 14th. 1AI notes that POINTS1.5 still follows the classic LLaVA architecture used in POINTS1.0, which consists of a vision encoder, a projector, and a large language model. According to the official introduction, this generation of POINTS model not only takes into account the efficiency-first idea insisted in POINTS1.0, but also extremely...

Tencent announced on December 3, mixed yuan large model on-line and open source literate video capabilities, the number of participants 13 billion, support for bilingual input in English and Chinese. Officials claimed that Tencent hybrid video generation model can generate "ultra-realistic" high-quality video, generated by the screen is not easy to deform; in the mirror or mirror scene, you can do the mirror reflection action and the outside completely synchronized, light and shadow reflections basically in line with the laws of physics. According to reports, Tencent hybrid video generation model using DiT architecture, adapted to a new generation of text encoder to improve semantic compliance, better response to multiple subjects depicted, to achieve more detailed finger...

Tencent's public relations director Zhang Jun announced on the evening of November 28, Tencent mixing big model December 3 officially on-line video generation capabilities, users can submit an application for trial in the Tencent Yuanbao App. Enterprise customers through the Tencent cloud to provide service access, the current API synchronization open internal test applications. At the same time, Tencent will also open source the video generation big model, the number of participants 13 billion. In the middle of this month, Tencent large model App Yuanbao released version 2.0, interface upgrades include updating the dialog list, support for the historical use of asset deposition, users can more easily find the history of...

November 19th, according to AI front reported today, Tencent outstanding scientist, Tencent hybrid big model technology leader Liu Wei, has left Tencent. In May this year, Liu Wei attended the AICon Global Artificial Intelligence Development and Application Conference organized by InfoQ in the capacity of one of the technical leaders of the hybrid model and delivered a keynote speech on "Tencent's hybrid model technology and application practice". In September, he also delivered a keynote speech titled "The Development of Generative AI and the Practice of Tencent's Hybrid Big Model" at the 2024 Bund New Media Conference...

Tencent's official public post announced that its AI intelligent workbench ima is now officially on line, in addition to completing the functions of the rest of the big model AI Q&A products (note: Q&A, create text, generate pictures, etc.), but also through the WeChat public number article ecology, the entire public number world of knowledge can be used by the user. The official claim is that users can find high-quality answers to all industry questions, from getting started to becoming proficient, through ima. ima's system model will comprehensively retrieve the entire network of public number sources according to the user's query, combining the matching degree, freshness, and other factors to...

Learned from Tencent Mixed Yuan, Tencent Yuanbao 2.0 version is officially launched today, bringing upgrades to the interface, model, experience and other aspects. According to reports, the upgrades to the interface include updating the dialog list to support the historical use of asset deposition, so users can more easily find the historical dialog and used intelligent body; new AI application exclusive plate, selected AI search, AI reading, creative painting, inspiration gallery, AI beauty photo five applications, and the "application plaza" provides a rich variety of types of intelligent body. In addition, the AI search function is not only available on WeChat, but also on WeChat's website. In addition, the AI search function has upgraded and integrated WeChat Video, QQ Music and other letters in addition to WeChat Public...

Tencent Holdings today released its financial results for the third quarter of 2024 ended September 30: revenue was 167.2 billion yuan, up 8% year-on-year, and net profit was 53.2 billion yuan, up 47% year-on-year. Non-IFRS net profit was 59.8 billion yuan, up 33% year-on-year. According to a report by Sina Science and Technology, Citi analysts asked the question "Will Tencent's big model app Yuan Yuan eventually embed commercial sponsorships" during the earnings call after the release. conference call asked, "Tencent big model App Yuan Yuan will eventually embedded commercial sponsorship answer?" Tencent Chief Strategy Officer James Mitchell said...

Tencent announced the launch of the Hunyuan-Large model, which is the largest Transformer-based MoE model that has been open sourced in the industry, with 389 billion total parameters (389B) and 52 billion activation parameters (52B). Tencent has open sourced Hunyuan-A52B-Pretrain, Hunyuan-A52B-Instruct, and Hunyuan-A52B-Instruct-FP8 at Hugging Face, and released...

Tencent announced on October 25, in order to help keepers observe pandas in all aspects, Tencent, the China Conservation and Research Center for the Giant Panda, Guangdong University of Technology, jointly created the world's first panda behavioral intelligent identification model and intelligence system. The model can identify the daily behavior of pandas such as eating, drinking, sleeping, and automatically generate daily, weekly, monthly reports and other visual data reports. By optimizing the SlowFast algorithm, the project team has significantly improved the system's ability to recognize behaviors in sheltered environments, and the accuracy of panda behavior recognition in sheltered indoor scenarios has increased to more than 80%. ...

China Academy of Information and Communications Technology (hereinafter referred to as "CIICT") and Tencent signed an AI business cooperation agreement on October 12th. The signing of this cooperation agreement is aimed at pooling and integrating the advantageous resources of the two sides to carry out all-round cooperation in technology innovation, standard development, assessment and evaluation, application demonstration, industrialization promotion, and consulting and exchange on the basis of resource sharing and mutual benefit and win-win situation, focusing on the key directions of AI cutting-edge modeling, software-hard collaboration, data resources, AI+, security governance, and international cooperation. Jiang Jie, vice president of Tencent Group, revealed that Tencent will take the signing of this cooperation as an opportunity to work with...

Tencent has officially learned that at the 2024 World Design Capital Conference, Tencent and Tongji University announced a cooperation, in which the two sides will combine Tencent's hybrid big model capabilities with Tongji University's strengths and influence in the discipline of big design, to jointly build the "Tongji-Tencent Big Design Big Model Initiative (TT)". The purpose of this initiative is to create the first design big model ecology in the design field that connects disciplines and industries; connects design schools, experts, industries and consumers; and helps realize large-scale and personalized teaching, research and industrial innovation. The initial attempt of the cooperation between the two sides is the "University Professor AI Intelligent Body" Prof....

Tencent Hybrid recently announced that its AI intelligent body product "Tencent MetaWare" now supports the release of WeChat public numbers, bringing a number of new features to public number operators. These features include: Real-time interaction with digital doppelgangers: public number operators can create digital doppelgangers to interact with fans in real time, improving user engagement and experience. 24/7 Intelligent Customer Service: Intelligent body can provide 7*24 hours customer service, such as checking express delivery, new student guidance, etc., to improve service efficiency. Article Insertion Function:Intelligent body can insert relevant content into public number articles to enhance the interactivity and information of the content. Provide readers with Q&A assistant, article more...

On September 5, Tencent announced the launch of a new generation of large models "Hunyuan Turbo". Compared with the previous generation model, Tencent Hunyuan Turbo has significantly improved performance, with training efficiency increased by 108%, inference efficiency increased by 100%, inference cost reduced by 50%, and decoding speed increased by 20%. The effect is comparable to GPT-4o in multiple benchmarks, and third-party evaluation ranks first in China. At the Tencent Global Digital Ecosystem Conference, Qiu Yuepeng, Vice President of Tencent, COO of the Cloud and Smart Industry Business Group, and President of Tencent Cloud, announced that Tencent Hunyuan Turbo was launched on Tencent Cloud, and the input and output prices are only half of the previous model...

Tencent HunyuanDiT recently released three new controllable plug-ins ControlNet with the community, namely tile (high-definition magnification), inpainting (image restoration and expansion) and lineart (line draft raw image), further expanding its ControlNet matrix. The addition of these plug-ins enables the HunyuanDiT model to cover a wider range of application scenarios, including art, creativity, architecture, photography, beauty and e-commerce, etc. 80% cases and scenarios provide more accurate...

Tencent recently launched an educational auxiliary tool, "AI Composition Tutoring", which aims to help students improve their writing skills. This feature is based on Tencent's Hunyuan Model and is now available for free on QQ Browser. The "AI Composition Tutoring" function covers all grades from elementary school to high school, and is synchronized with the teaching focus of each grade to provide students with personalized composition tutoring. Users only need to enter the composition title, and AI can output the idea, outline, materials and model essays step by step to help students build writing logic. In junior high and high school, where the genre classification is more clear, students can also choose argumentative essays or narrative essays for targeted tutoring. …

On August 9, the Chinese Institute of Electronics (CIE) and Tencent jointly launched the "CIE-Tencent Doctoral Research Incentive Program (Hybrid Large Model Special)". The program provides financial and research platform support to doctoral students with strong interest and clear goals in big model research. Scope of Attached Specialties Computer Science, Artificial Intelligence, Software Engineering, Big Data and other related majors are eligible to apply. Application Requirements Chinese doctoral students enrolled in universities and research institutes at home and abroad who are members of the Chinese Institute of Electronics (CIE) and meet the following conditions are eligible to apply. PhD graduation no earlier than July 2026; Take the first...

Tencent Yuanbao announced that in the latest version of Tencent Yuanbao, when users upload URL links or documents of professional content such as essays, financial reports, research reports, etc., in addition to obtaining a textual overview and summary, they can also enter an in-depth reading mode to read the long article in depth. This mode provides an overview of the core content and modularized analysis, generating summary charts to assist users in quickly understanding the key information. According to reports, the depth of reading mode can natively support the input of up to nearly 500,000 words, can be in the understanding of professional content based on the generation of illustrated content, not only can quote the input content of the original map, but also through the code to draw a higher-order analysis of...

It is reported that Tencent participated in the financing of Moonshot AI, a domestic artificial intelligence unicorn and Kimi smart assistant developer, with a $300 million (currently about RMB 2.151 billion) round. In response, Moonshot AI responded to the reporter of The Paper: "No comment on the financing news." IT Home noted that in February this year, it was reported that Moonshot AI had completed a new round of financing of more than $1 billion (currently about RMB 7.171 billion), with investors including Sequoia China, Xiaohongshu, Meituan, Alibaba, and old shareholders following suit. After this round of financing, Moonshot AI is estimated to be…

Tencent Cloud Zhihui is an intelligent material design and production platform dedicated to providing convenient, efficient, and high-quality material production services for individual creators and small and medium-sized enterprises. Based on Tencent's self-developed AI intelligent drawing engine, AI intelligent video synthesis engine, AI color matching algorithm, intelligent typesetting engine and other technologies, Tencent Cloud Zhihui can realize the intelligent design and production of materials to meet the creative needs of various types and industries. Through Tencent's big data processing, intelligent recommendation engine, voice processing and other technologies, users can obtain more accurate creative recommendations, faster production speeds, and more stable and reliable production quality. Tencent Cloud Zhihui is...

Tencent AI Lab launched TransAgents, a translation multi-agent framework specifically for translating long literary content, providing translations in multiple languages including Chinese, English, Japanese, and Korean. TransAgents GitHub address: https://github.com/minghao-wu/transagents Paper address: https://arxiv.org/pdf/2405.11804 TransAgents is a "virtual translation company" that mimics human social...

Tencent Sogou Input Method has launched new AI functions today, including AI writing, AI dialogue, quick Q&A, AI pets, AI selfie expressions and other functions. AI writing claims to “write a title in 1 second and produce the full text in 10 seconds”, and provides 21 writing commands, including chat polishing, all-around writing, shopping reviews, speech generation, short video dynamics, poetry for you, book and movie reviews, email writing, highly praised friend circle, personal signature, seed notes, work reports, creative blog posts, knowledge sharing, resume polishing, self-introduction, exam wishes, hot titles, great reviews, job hunting words…

Tencent Hunyuan Wenshengtu Large Model (Hunyuan DiT Model) announced the full open source training code, including LoRA plug-in and ControlNet plug-in. LoRA is a technology for fine-tuning large language models. It can train a model with specific features using a small amount of data without modifying the original model or increasing the model size. ControlNet is a controllable generation algorithm that allows users to better control image generation by adding additional conditions. Tencent Hunyuan provides three first-run ControlNet models that can extract application-related...

Tencent recently announced that its AI big model application "Tencent Yuanbao" has completed the upgrade of AI search and parsing capabilities. The new version of "Tencent Yuanbao" supports processing a single document up to 10 million words, parsing up to 50 files at the same time, with an upper limit of 100MB. supported file formats include PDF, DOC, TXT, XLSX, PPTX, etc. In addition, the application can quickly generate bar charts, line graphs, line graphs, and PPTX based on the parsed content. In addition, the application can also quickly generate bar charts, line graphs, pie charts and other data charts based on the parsed content. Tencent Yuanbao's upgrade also includes the ability to recognize and parse image formats...

Recently, Tencent revised the "Implementation Rules for Video Account Showcase Experts to "Publish Low-Quality Content", and the revised content was solicited from June 7 to June 13 this year. The latest plan shows that "using plug-ins, AI and other tools to generate virtual images for live broadcasts", "providing, selling, teaching or displaying content that the platform does not advocate: such as explaining and selling virtual human broadcast software", etc., involving virtual human/digital human live broadcasts, will be clearly listed as low-quality content by the video account and will be handled as violations. According to a report by The Paper today, relevant sources revealed that using digital human/virtual human live broadcasts...