Tencent Hunyuan

On January 27th, a message was sent to the Mixer (January 26) to publish a Mixed Image 3.0-Instruct model, which supports image editing and multi-mixing, and goes online in parallel with the Equestrian Treasure, where users can experience it directly at the end of the Mixer and at the Quantuan Network. Mixtures Image 3.0 Total Parameters of Maps 8 billion (Activating Parameters around 1.3 billion), using a hybrid specialist (MoE) structure, based on the original multi-modular structure basic model of Mixtures Image 3.0, introducing multi-task data of the Graphics through..

On December 31, in the news of the announcement of the open-source translation model 1.5, the text consists of two models: Tencent-HY-MT1.5-1.8B and Tencent-HY-MT1.5-7B, in support of 33 language translations and 5 Chinese/linguistics, in addition to common languages such as Chinese, English and Japanese, as well as small languages such as Czech, Marathi, Estonian and Icelandic. Now both models are placing on the M.O. Online, and open-source communities like Github and Huggingface..

ON NOVEMBER 27TH, THE 3D CREATIVE ENGINE WAS OFFICIALLY LAUNCHED AND IS OPEN TO INTERNATIONAL USERS. AT THE SAME TIME, THE HYBRID 3D MODEL API IS ONLINE AT THE INTERNATIONAL STATION OF TELECOMMUNICATION CLOUDS, THROUGH WHICH DEVELOPERS AND BUSINESS USERS CAN ACCESS THE MODEL CAPABILITY QUICKLY. THE 3D MODEL IS KNOWN TO BE A LARGE-SCALE AI MODEL FOR SELF-INTELLECTION AND ONE OF THE MOST POWERFUL MODELS IN THE SERIES, WITH A GOOD REPUTATION AMONG DOMESTIC AND FOREIGN DEVELOPERS AND 3D CREATORS. SO FAR, THE 3D MODEL COMMUNITY DOWNLOADS OVER 3 MILLION, THE WORLD'S MOST POPULAR..

On November 25th, a new open source model, the HunyuanOCR, was announced today, with only 1B, based on a multi-modular structure of the hybrids, with many industry OCR application lists SOTA (note: state of the art). According to official sources, thanks to the conceptual design of the Magnificent Multi-Mode Model “end-to-end” philosophies, the functions of the HunyuanOCR can achieve optimal results only if it is based on a single forward reasoning. The OCR expert model is built on a hybrid multimodular structure and consists of three main components: native..

On November 13th, the official public call for information announced that Liblibai, the country ' s largest AI content-creation platform, had come on line today to get the mixed image 3.0 (HunyuanImage 3.0) and the YouTube-Video FX video effects model for the euphoria. Users can use the model through a photo video generator on the LiblibAI platform. According to the official presentation, the 3.0 model, with a parameter size of 80B, is the first open source industrial-grade multimodular model, which is..

On September 16th, a 3D 3.0 AI model was officially released today, with a three-fold increase in modelling accuracy and geometric resolution of 15363, supporting 3.6 billion bodyware modelling. The first 3D 3.0 3D-DIT sculptor model, which claims that it is difficult to break traditional technology to balance the bottlenecks of overall structure, sequencing and local detail, and to address the two hard points of generation of overall structure and local detail, “to achieve a sharp edge, well-defined super-high 3D geometry” and to maintain the industry SOTA. Compared to previous generations..

Messages of September 12, the latest day, the Twitter search interface was updated, the Twitter AI search entered the first level of the portal, and the relevant buttons were visible by clicking on the search box at the top of the first page. It is divided into three sections: deep thinking, uploading pictures, uploading files. In addition to the Deep Thinking option, which allows for the DeepSeek-R1 model or the Quest T1 model, the user can choose the quick answer option to provide the most common answer. The uploading of pictures or taking of photographs can be asked on the basis of the pictures and support for such functions as graphics, question-taking, search of goods, etc. Uploading a file or a public sign can summarize..

September 1 news, Tencent mixed yuan announced its international translation model open source, for developers to download and deploy for free. It is understood that this model is named Hunyuan-MT-7B, the total number of references is only 7B, supports 33 languages, 5 kinds of Chinese language / dialect translation, is a lightweight translation model with comprehensive capabilities. According to Tencent Hybrid, in the International Association for Computational Linguistics (ACL) WMT2025 competition that ended at the end of August, Tencent Hybrid's Hunyuan-MT-7B (Entry Name: Shy-hunyuan-MT) took all the...

August 28 news, Tencent mixed yuan today announced open source end-to-end video sound generation model Hunyuan-Foley, users only need to enter the video and text, you can match the video with movie-level sound. According to the official introduction, HunyuanVideo-Foley not only breaks the limitations of AI-generated videos that can only be "watched" but not "listened to", making silent AI videos a history, but also truly achieves the following Foley not only breaks the limitations of AI-generated videos that can only be "seen" but not "heard", but also makes silent AI videos a thing of the past, and truly realizes the immersive audio-visual experience of "reading the picture, reading the text, and matching the sound. This sound generation tool can be widely used in short video creation, movie...

Yesterday, someone said, "I don't have enough free points for Dreams, what other AI tools can I use to make short videos for free?" AI tools update faster than the book, the old is not enough, the new must be arranged. Directly on the dry goods, 5 AI video tools, there is always a solution to your urgent needs! Baidu drawing think Baidu launched the AI video creation platform, Xiao Yi also wrote an article before introduced: free public beta, no watermark download, Baidu AI video kill crazy! Advantages: A sentence summarizes the quality of quality, currently free, download no watermark, no points limit, and use and cherish. Deficiencies: The purpose...

August 4 news, Tencent mixed yuan today released four open source small size model, parameters were 0.5B, 1.8B, 4B, 7B, consumer-grade graphics cards can run, applicable to laptops, cell phones, smart cockpit, smart home and other low-power scenarios, and support for vertical areas of low-cost fine-tuning. Currently, all four models are online in open source communities such as Github and Huggingface, and several consumer-grade terminal chip platforms such as Arm, Qualcomm, Intel, and MediaTek have announced their support for deployment. The four newly open-sourced models are fusion inference models with ...

May 20, 2011 - Tencent's Mixed Meta team just announced the official release of Tencent's Mixed Meta Game Visual Generation Platform (hereinafter referred to as "Mixed Meta Game"). This is the first industrial-grade AIGC game content production engine, which significantly optimizes the game asset generation and game production process. According to the official introduction, Tencent Hunyuan positioned Hunyuan-Game as the first AIGC series of large models for industrial-grade game content asset generation. The technical report shows that it covers a complete game image and video modeling capability matrix, including four major game image models and five major video models. The four image generation...

Tencent's Hunyuan Image 2.0 model was officially released on May 16th, and is available on Tencent's official Hunyuan website and open to the public for registration and experience. Tencent's Hunyuan Image 2.0 model has two main features: real-time image generation and ultra-realistic image quality. Compared with its predecessor, Tencent's hybrid image 2.0 model has increased the number of parameters by an order of magnitude, and thanks to the ultra-high compression multiplier image codec and a new diffusion architecture, the speed of image generation is "significantly faster than the industry's leading models", and the inference speed of each image in similar commercial products takes 5 to 10 seconds! ...

May 9 news, tencent mixed yuan announced today officially launched and open source new multimodal customized video generation tool Hunyuan Custom. the model is based on mixed yuan video generation big model (Hunyuan Video) to build, the body consistency effect than the existing open source program. According to reports, Hunyuan Custom integrates the ability of text, image, audio, video and other multimodal inputs to generate video, is an intelligent video creation tool with a high degree of control and generation quality. Tencent said that Hunyuan Custom model...

April 23 news, Tencent today announced that the hybrid 3D generation model new upgrade, v2.5 new version in the modeling of the fineness of a substantial increase. The hybrid 3D AI creation engine has been fully updated to the v2.5 model base, while the free generation quota has been doubled to 20 times per day. The hybrid 3D generation API has also been launched on Tencent Cloud and is open to enterprises and developers. Compared to its predecessor, the hybrid 3D v2.5 model architecture has been upgraded, with the number of total parameters raised from 1B to 10B, the number of effective facets increased by more than 10 times, and the effective geometric resolution...

1AI learned from Tencent mixed yuan WeChat public number, March 21 Tencent mixed yuan officially launched the official version of the self-developed deep thinking model mixed yuan T1. According to the official introduction, "T1" spits out words fast, can return in seconds, but also good at ultra-long text processing, is Tencent's self-research strong reasoning model. Through large-scale reinforcement learning, and combined with special optimization of mathematics, logical reasoning, science and code and other science challenges, the official version of Mixed Elements T1 further improves its reasoning ability. On common benchmarks that reflect the basic capabilities of inference models, such as the large language model evaluation enhancement dataset MMLU-PRO,...

On March 19, Tencent announced via its official public website that its new reasoning model, T1, will be officially released this Friday, March 21 at 23:00 BST. At the same time, Tencent announced that for the first time, the hybrid model has made it onto the Chatbot Arena list, ranking among the top 15 in the world, where users anonymously interact with multiple models to vote on which is better, generating a leaderboard based on their scores. This kind of evaluation is also seen as a direct PK arena for big models, which is simple and straightforward. On September 5 last year, Tencent...

March 18th, in today's Tencent Mixed 3D Open Source Day, Tencent Mixed announced the launch of five new 3D generation models, which have been improved in generation speed, detail and material expression, and are all open source. At the same time, its self-developed 3D AI creation engine has also been upgraded, with new capabilities such as multi-view input, intelligent model surface reduction, and full format compatibility, and is fully open for use by C-support users. The 5 open-source models are all based on Hunyuan3D-2.0: Turbo Series Models: On the basis of guaranteeing high precision and quality, the...

On March 6, Tencent hybrid released and open-sourced a graph-generated video model, as well as on-line audio and action-driven capabilities. Tencent mixed yuan released and open source graph born video model: can generate 5 seconds of short video, but also automatically with background sound effects After seeing, decided to try Tencent mixed yuan action-driven ability effect how? Below, together to learn how to hybrid AI video to generate similar template action video? ▌How to generate similar template action video with hybrid AI video? Go to the official website of Mixmaster AI Video and click [Action Driver]. Go to the Action Driver page and click [Upload Character Picture]. Pothead here to select the previously made...

March 6 news, 1AI from tencent mixed yuan weibo public number was informed that tencent mixed yuan released graphic video model and open source, at the same time on-line lip-synching and action-driven gameplay, and support for the generation of background sound effects and 2K high-quality video. Based on the ability of graphic video, users only need to upload a picture and briefly describe how they want the screen to move, how the camera is dispatched, etc., and the hybrid can make the picture move according to the requirements and turn it into a 5-second short video, which can also be automatically matched with background sound effects. In addition, by uploading a picture of a character and typing in the text or audio you want to "lip-sync", the character in the picture can "talk"...

For public number operators, aggregating past articles into a knowledge base not only enriches the public number's content ecosystem, but also greatly improves the interactive experience with readers. Imagine how exciting it would be if the public number could act like a smart library! Automatically receive questions from readers and find relevant information from previously published articles to give accurate answers Now this operation is no longer a distant dream, but can be realized in less than a minute through the combination of DeepSeek and Tencent Hybrid. This article will show you how to public...

January 21 news, Tencent today announced that the open source on-line mixed yuan 3D generate big model 2.0. tencent mixed yuan also synchronized on-line mixed yuan 3D AI creation engine, said to be "the industry's first one-stop 3D content AI creation platform". The technology claims that a sentence, a picture, or even a sketch can generate a 3D model, and it can even add movements, change textures, pinch characters, and do animation. Tencent's Mixed 3D-2.0 version mainly upgrades the geometry and texture models in the 3D generation process. The task of the geometry model is to capture 3...

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...

According to Tencent Technology’s report today, the August list of the Chinese multimodal big model SuperCLUE-V benchmark was released, and Tencent’s Hunyuan big model ranked first among domestic big models (71.95 points). Tencent Technology claims that the model accurately identifies image elements and generates natural language descriptions, fully understanding and insight into details. This evaluation covered 12 highly representative multimodal understanding big models at home and abroad, and Tencent’s Hunyuan model scored 71.95 in multimodal basic capabilities and application capabilities. According to inquiries, the August list covers the 12 most representative domestic and foreign…