Latest AI Resources

AnyI2V is an image animation generation framework jointly launched by Fudan University, Alibaba Dharma Institute and others, which supports the conversion of static conditional images (e.g., grids, point clouds, etc.) into dynamic videos without the need for complex training processes and large amounts of data.

SRPO (Semantic Relative Preference Optimization) is a text-to-image generation model introduced by Tencent Hybrid, which optimizes the reward mechanism through text conditioned signals to achieve online adjustment of rewards and reduce offline fine-tuning dependency.

Qwen3-Next is a new generation of hybrid architecture big model open source by Ali Tongyi, combining Gated DeltaNet and Gated Attention technology, good at dealing with long text, fast inference and saving computing resources.

Wenxin Big Model X1.1 is a deep thinking model launched by Baidu, based on a hybrid reinforcement learning framework that focuses on improving language understanding and generation. The model excels in handling complex questions, following instructions and simulating the behavior of intelligences, and can accurately provide knowledgeable answers and high-quality text content.

HunyuanImage 2.1 is Tencent's open source graphic model, designed for high-quality image generation. The model supports native 2K resolution, can accurately render complex scenes and details, so that the character's expression and movement can be vividly reproduced.

AntSK FileChunk is a free intelligent document slicing tool designed for RAG (Retrieval Augmented Generation) applications. Semantic as the core, the document will be intelligently sliced into semantically complete, coherent segments , support for multi-language , can dynamically adjust the size of the slice to ensure that the context of coherence.

UnifiedTTS is a one-stop platform for text-to-speech (TTS) services. It supports multiple languages, including Chinese, English, Japanese and Korean, to meet the needs of global business. Through a unified API interface, it integrates many mainstream TTS services, including Micro...

MiniCPM 4.1 is an ultra-efficient end-side large language model introduced by Facade Intelligence. With InfLLM v2 sparse attention architecture, each lexeme only needs to calculate the relevance to less than 5% lexemes, which significantly reduces the processing overhead of long text. In a 128K long text scenario...

WeKnora is Tencent WeChat team open source based on the Large Language Model (LLM) document understanding and semantic retrieval framework , designed for the structure of complex, heterogeneous document content scenarios and designed to use a modularized architecture , integration of multimodal preprocessing , semantic vector indexing , intelligent recall and large model generative reasoning ...

XTuner V1 is a new generation of large model training engine open-sourced by Shanghai Artificial Intelligence Laboratory (SAL), designed for ultra-large scale sparse Mixed Expert (MoE) model training. Developed based on PyTorch FSDP, it achieves high performance through multi-dimensional optimization of memory, communication and load ...

Qwen3-ASR-Flash is Alibaba's latest high-precision speech recognition model, based on the Qwen3 base model, trained on massive multimodal data. It supports 11 languages and multiple accents, including Mandarin, Sichuan, Minnan, Wu, Cantonese and other dialects...

Seedream 4.0 is an advanced image generation and editing tool launched by ByteDance, centered on the integration of generation and editing, with powerful features such as precise command editing, high feature retention, and deep intent understanding.

rStar2-Agent is an advanced AI mathematical reasoning model open-sourced by Microsoft that demonstrates strong mathematical problem solving capabilities by achieving an accuracy of 80.61 TP3T in the AIME24 test. The model is equipped with scientific reasoning capabilities, achieving in the GPQA-Diamond benchmark...

Qwen3-Max-Preview is the latest flagship large language model released by Tongyi Qianwen. It is the model with the largest number of parameters in the Qwen3 family, with a parameter size of over 1 trillion. The model has significant improvements in inference, instruction following, multi-language support and long-tail knowledge coverage...

OneCAT is a new unified multimodal model launched by Meituan in conjunction with Shanghai Jiaotong University, which adopts a pure decoder architecture and can seamlessly integrate multimodal comprehension, text-to-image generation and image editing functions. The model abandons the design of traditional multimodal models that rely on external visual coders and disambiguators through modality-specific...

Claudable is an open source web application builder based on Next.js that combines the advanced AI agent capabilities of Claude Code and Cursor CLI with Lovable's simple and intuitive application building experience....

FineVision is Hugging Face's open source visual language dataset for training advanced visual language models. It contains 17.3 million images, 24.3 million samples, 88.9 million rounds of dialog, and 9.5 billion answer tokens. The dataset aggregates...

InfinityHuman is a commercial-grade long time-series audio-driven character video generation model jointly launched by ByteDance and Zhejiang University. The model is audio-driven and can generate high-resolution, long duration and visually consistent character videos.

HunyuanWorld-Voyager (Hunyuan Voyager for short) is the industry's first ultra-long roaming world model released by Tencent that supports native 3D reconstruction. It is a novel video diffusion framework that generates a 3D point cloud sequence of user-defined camera paths from a single image, supporting...

Hunyuan-MT-7B is a lightweight translation model introduced by Tencent's Mixed Meta Team, with 7 billion references, supporting the mutual translation of 33 languages and 5 folk-Chinese languages/dialects, including Cantonese, Uyghur, and Tibetan. In the International Association for Computational Linguistics (ACL) WMT2025 competition...

Step-Audio 2 mini is an open source end-to-end speech grand model of Step-Audio. It breaks through the traditional speech model structure and adopts the true end-to-end multimodal architecture, which directly transforms the original audio input into speech response output with lower latency, and understands paralinguistic information and non-vocal signals.

MobileCLIP2 is an upgraded version of MobileCLIP, an efficient end-side multimodal model introduced by Apple researchers. It is optimized in terms of multimodal reinforcement training by training better-performing CLIP instructor model integration on DFN datasets and improved graphical raw...

InternVL3.5 (Shusheng-Wanxiang 3.5) is an open source multimodal large model of the Shanghai Artificial Intelligence Laboratory, the model is fully upgraded in terms of general ability, reasoning ability and deployment efficiency, providing nine sizes of versions from 1 billion to 241 billion parameters, covering different resource demand scenarios, including thick...

FastVLM (Fast Vision Language Model) is an efficient visual language model introduced by Apple Inc. With FastViTHD hybrid visual coder as the core, it incorporates convolutional and Transformer architectures to significantly reduce visual...

Meeseeks is an open source large model evaluation set used by the Meituan M17 team to evaluate the model's ability to follow instructions.Meeseeks uses a three-tiered evaluation framework to comprehensively measure whether the model is able to generate answers in strict accordance with the user's instructions from the macro to the micro level, without evaluating the knowledge of the content of the answers positively ...

gpt-realtime is an advanced speech model from OpenAI that supports direct audio processing to generate natural and smooth speech. The model supports multiple languages and styles, understands non-verbal cues such as laughter, and can switch between languages.

Youtu-agent is an open source framework for building and running autonomous intelligences from Tencent Youtu Labs. The framework performs well in WebWalkerQA and GAIA benchmarks, with an accuracy of 71.47% and 72.8% respectively.The framework...

HunyuanVideo-Foley is an open source video sound generation model by the Tencent Mixed Yuan team that supports adding accurately matched sound effects to silent videos. The model is based on a large-scale dataset training , with a multimodal diffusion transformer architecture , combined with the characterization of the alignment loss function and audio VAE optimization techniques ...

PixVerse V5 is a big model of AI video generation launched by Aishi Technology. The model can generate high-quality video content based on user-input text descriptions or images, and supports multiple styles, such as anime, sci-fi, and national style.

Ask White 5 is the flagship "All in One" model with a very high level of intelligence. The model has excellent performance in many assessments, such as the AA-Index composite assessment score of 64.7 and the STEM ability assessment score of 86, which is close to the world's leading GPT-5.

MiniCPM-V 4.5 is an open source 8B parametric multimodal model of Facade Intelligence, built based on Qwen3-8B and SigLIP2-400M, with the ability to efficiently process images and videos. It has excellent performance in visual token consumption, processing ...

Gemini 2.5 Flash Image (codename nano banana) is a state-of-the-art image generation and editing model from Google that maintains the consistency of characters across different scenes and supports precise image editing through natural language, such as blurring backgrounds and removing stains.

Wan2.2-S2V is Ali Tongyi open source multimodal video generation model , only a static picture and a piece of audio , you can generate high-quality digital human video , and supports a variety of image types and frame .

ChatGPT Tip Engineering for Developers is a joint DeepLearning.AI and OpenAI course designed for developers, featuring Isa Fulford, Andrew Ng to teach how to use Large Language Models (LLMs...

Ask White o4 is an innovative parallel thinking model that opens 8 thinking paths at the same time, analyzes the problem from multiple perspectives and automatically filters out the optimal solution. The model incorporates advanced Long-CoT reinforcement learning and process reward learning techniques, has powerful deep reasoning capabilities, and performs well in complex tasks.

VibeVoice is a new text-to-speech (TTS) model from Microsoft. The model generates conversational audio from up to four different speakers and supports up to 90 minutes of continuous voice output, breaking the length limitations of traditional TTS systems.

SpatialGen is an open source 3D scene generation model of Qunar Technology, based on the diffusion model architecture, supporting the generation of spatio-temporally consistent multi-view images based on textual descriptions, reference images and 3D spatial layouts, and further generating 3D Gaussian scenes and rendering roaming videos.

EchoMimicV3 is a multimodal digital human video generation model introduced by Ant Group, with 1.3 billion parameters, capable of handling multiple inputs such as audio, text, images, etc. to generate high-quality digital human animations.

Fun-ASR is a big model of speech recognition jointly launched by Nail and Tongyi Labs. The model has been trained with massive audio data and can accurately recognize multi-industry terminology, such as Internet, technology, home decoration, etc., significantly improving the recognition accuracy. The model combines with Nail enterprise information for inference optimization to reduce the illusion problem...

Squibler is a powerful AI-assisted writing platform designed for writers that helps users with the entire process from conception to creation to publication. The platform provides a variety of story templates covering novels, screenplays, short stories, etc. Users only need to enter the initial concept, and the AI can generate outlines, characters, scenes...

91 Writing is a fully open source AI novel creation tool, developed based on Vue 3 and Element Plus, integrating a variety of advanced AI models, such as GPT, Claude, Gemini, and so on. The tool provides creators with a complete creation tool chain from idea to text, including project creation...

Aivilization is the world's first AI multi-intelligent body social simulation platform developed by the Hong Kong University of Science and Technology. It builds a visual digital sandbox where users can create and guide thousands of AI intelligences to observe the social evolution of future human-AI coexistence. The platform supports...

Grok 2.5 is an open source AI model from Elon Musk's xAI. With 269 billion parameters, it is based on the Mixed Expert (MoE) architecture for powerful performance and inference. The model has been tested at graduate level scientific knowledge (GPQA), generalized knowledge (MMLU, MM...

Draw A Fish is simple and fun online AI fish drawing site where users can draw fish patterns and place them in a globally shared virtual fish tank.Draw A Fish requires no registration and is easy to use, taking only seconds to create and share.

MIT's latest report, The Generative AI Divide: the State of Business AI in 2025, reveals the core of the generative AI (GenAI) adoption process that companies are experiencing through in-depth research of more than 300 AI projects, interviews with 52 organizations, and a survey of 153 executives...

AutoClip is open source AI video editing tool, based on advanced AI technology to realize the whole process of automated video processing. Tools can automatically identify the highlights of the video, accurate extraction of valuable content, can be based on the similarity of the theme of intelligent clustering, to generate a collection of content.AutoClip support...

ToonComposer is a generative AI animation tool jointly launched by The Chinese University of Hong Kong, Tencent PCG ARC Lab and Peking University. Through generative post keyframe technology, the intermediate frame generation and coloring process is integrated into an automated process, requiring only a sketch and a...

Seed-OSS is a large family of language models open-sourced by the Byte Jump Seed team, focusing on long text and reasoning tasks. The model performs well in complex logical reasoning and multi-step reasoning, with high accuracy and efficient problem solving.Seed-OSS supports long text contexts up to 512K...

Intern-S1-mini is a lightweight scientific multimodal macromodel with parameter scale of 8B launched by Shanghai Artificial Intelligence Laboratory (SAL).It inherits the powerful capabilities of Intern-S1, combining both general and specialized scientific capabilities, and is suitable for rapid deployment and secondary development. In terms of performance, I...

Klear-Reasoner is a high-performance inference model from Racer, based on Qwen3-8B-Base. The model is trained by long thought chain supervised fine-tuning and reinforcement learning to perform well in mathematical and code reasoning.Klear-Reasoner...

CombatVLA is an innovative 3D action role-playing game (ARPG)-specific model from the Future Life Lab team of the Amoy Sky Group.CombatVLA is a visual-linguistic-action (VLA) model, built on a 3B parametric scale, that collects human player's through a motion tracker...

DeepSeek V3.1 is a new generation of AI models introduced by DeepSeek, with important upgrades based on its predecessor, V3. DeepSeek V3.1 introduces a hybrid reasoning architecture that allows the model to flexibly switch between thinking and non-thinking modes, significantly improving the thinking...

Qwen-Image-Edit is an all-purpose image editing model introduced by Ali Tongyi, built on the Qwen-Image architecture with 20 billion parameters. The model combines both semantic and appearance editing capabilities, and can perform low-level visual appearance editing on images (e.g., adding, deleting...

MoE-TTS is a speech synthesis framework introduced by KunlunWanwei, based on the Mixed Expert (MoE) architecture, which combines pre-trained Large Language Models (LLMs) with speech expert modules.MoE-TTS retains the powerful textual reasoning by freezing the textual module parameters and updating only the speech module parameters...

Genie Envisioner (GE) is a unified platform for robot operation developed by the Genie Robotics team in collaboration with the National University of Singapore, Beijing University of Aeronautics and Astronautics and other organizations. It allows robots to better understand and perform tasks by "imagining first, then acting".

DINOv3 is a next-generation self-supervised vision base model from Meta AI, which adopts a self-supervised learning paradigm to learn image features without labeling data. It solves the feature degradation problem by improving data preparation and introducing Gram anchoring, and improves the generalization...

Mureka V7.5 is a state-of-the-art AI music generation model from Kunlun World Wide, focusing on Chinese songwriting. The model can accurately reproduce tones and playing techniques to generate natural, smooth and emotional vocals. Based on optimized automatic speech recognition (ASR) technology, Mureka V...

Skywork Deep Research Agent v2 is a deep research intelligent body launched by Kunlun Wave, focusing on the integration and analysis of multimodal information.Skywork Deep Research Agent v2 can process text, graph...

Hunyuan-GameCraft is Tencent Hunyuan team open source interactive game video generation framework. Framework from a single picture and prompts to generate highly dynamic game video , support the user through the keyboard and mouse to control the video content in real time .

RynnRCP is an open source Robot Context Protocol (RCP) from Ali Dharma Institute that lowers the threshold for development of embodied intelligence and opens up the entire development process.RynnRCP consists of the RCP framework and the RobotMotion module.The RCP framework, through capability abstraction and multi-protocol support, will...

RynnEC is a world understanding model introduced by Alibaba Dharma Institute, focusing on embodied intelligence tasks. The model is based on multimodal fusion technology, combining video data and natural language, and can parse objects in a scene from multiple dimensions, supporting functions such as object understanding, spatial perception and video target segmentation.

Matrix-3D is an open source framework from Skywork AI team, focusing on generating explorable panoramic 3D worlds. The framework combines panoramic video generation and 3D reconstruction techniques to generate high-quality, omni-directional explorable 3D worlds from a single image or text prompt...

GLM-4.5V is the world's leading open source visual inference model introduced by Smart Spectrum, with 106 billion total parameters and 12 billion activated parameters. The model is trained based on the new generation text base model GLM-4.5-Air, with powerful visual understanding and reasoning capabilities, capable of handling images, video...

Matrix-Game 2.0 is a self-developed interactive world model released by Kunlun SkyWork AI. Matrix-Game 2.0 is the industry's first open-source, real-time, long-sequence interactive generation model for general-purpose scenarios. The model is able to run at 25 FPS through a visually-driven interaction scheme in multiple...

Baichuan-M2 is an open source medical augmented large model launched by Baichuan Intelligence. It performs well in the medical field, especially in the HealthBench review with a score of 60.1, surpassing OpenAI's gpt-oss120b and many other open source models, becoming a global...

Qwen-Flash is a high-performance, low-cost language model introduced in the Alibaba Tongyi Thousand Questions series, designed for fast response and efficient processing of simple tasks. Based on the advanced Mixture-of-Experts (MoE) architecture, it is realized by sparse expert network...

SkyReels-A3 is an audio-driven digital human creation tool from Kunlun World Wide Group. SkyReels-A3 is an audio-driven digital human creation tool, which can generate high-quality dynamic video content through simple inputs (e.g., portrait images and voice), make static photos "come alive", and replace lines for existing videos with new lip-syncs that the characters will automatically...

dots.vlm1 is the first multimodal big model open-sourced by Little Red Book hi lab. Based on NaViT, a 1.2 billion parameter visual encoder trained from scratch, and DeepSeek V3 Large Language Model (LLM), it has powerful visual perception and text inference...

Genie 3 is a next-generation universal world model from Google DeepMind that enables the generation of highly dynamic and coherent virtual worlds in real time.Genie 3 simulates physical phenomena, natural ecosystems, and supports the creation of fantasy and historical scenarios. With text prompts, users can...

Claude Opus 4.1 is a state-of-the-art large-scale language model from Anthropic, designed for efficient processing of complex tasks. The model excels in the programming domain, generating high-quality code, supporting up to 32k of single output, and adapting to a wide range of programming styles...

gpt-oss is a family of open source inference models from OpenAI that enable efficient, flexible, and easy-to-deploy AI solutions for developers. gpt-oss consists of two versions, gpt-oss-120B with 117 billion parameters and support for 8...

MiDashengLM is Xiaomi's open source large model for efficient sound understanding, with specific parameter version MiDashengLM-7B , focusing on audio processing and understanding. The model is based on Xiaomi Dasheng audio encoder and Qwen2.5-Omn...

MOSS-TTSD is an open source spoken dialog speech generation model developed by the Speech and Language Laboratory of Tsinghua University. MOSS-TTSD can convert text dialog scripts into natural, smooth and expressive conversational speech, and supports bilingual generation in English and Chinese.

AudioGen-Omni is a multimodal audio generation model from Racer that generates high-quality audio, speech, and songs based on inputs such as video, text, etc.AudioGen-Omni is based on advanced techniques such as multimodal diffusionTransformer and phase-aligned...

LangExtract is a Google Open Source Python library that uses large language models (LLMs) to extract structured information from unstructured text. With user-defined commands and a handful of examples, it can efficiently identify and organize key details, such as clinical notes from...

FastDeploy is a high-performance reasoning and deployment tool from Baidu, designed for Large Language Models (LLMs) and Visual Language Models (VLMs).FastDeploy is developed based on the Flying Paddle (PaddlePaddle) framework, and supports a variety of hardware platforms...

InteriorGS is a high-quality 3D Gaussian semantic dataset introduced by Qunar Technology. The dataset contains 1,000 3D scenes covering more than 80 indoor environments such as homes, convenience stores, wedding halls and museums. The dataset has more than 554,000 object instances in 755 categories...

DragonV2.1 is an advanced zero-sample text-to-speech (TTS) model from Microsoft. Based on the Transformer architecture, the model supports multi-language and zero-sample speech cloning, and generates natural, expressive speech with only 5-90 seconds of voice prompts.

ScreenCoder is an open source intelligent tool to quickly convert UI design screenshots into high quality HTML/CSS code. Tools based on modular multi-intelligence architecture , combined with visual understanding , layout planning and code synthesis techniques to support the generation of high-precision and semantic front-end ...

Gemini 2.5 Deep Think is an AI reasoning model from Google designed to solve complex tasks. It is a variant of the model that won the gold medal at the International Mathematical Olympiad (IMO) 2025, and is designed to solve complex tasks through Parallel ...

MindLink is a large model of open source reasoning launched by Kunlun World Wide Web. With adaptive reasoning mechanism , according to the complexity of the task can be flexibly switched inference mode , simple tasks quickly generated , complex tasks in-depth reasoning , taking into account the efficiency and accuracy . Plan-driven reasoning paradigm to remove the "think" label , down ...

Kimi K2 High Speed Edition (kimi-k2-turbo-preview) is a high-performance language model introduced by Kimi, the Dark Side of the Moon. The model is optimized on the basis of Kimi K2, the output speed is greatly increased, and 40 Token per second can be generated...

dots.ocr is a multilingual document parsing model open-sourced by Xiaohongshu hi lab, based on a 1.7 billion-parameter visual language model (VLM), which can efficiently perform document layout detection and content recognition while maintaining a good reading order.

HYPIR is a large model for image restoration introduced by Dong Chao's team at Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences. The model combines the fractional prior of diffusion modeling with adversarial generative networks to achieve efficient, high-quality image restoration.HYPIR can quickly restore old photos and improve resolution while keeping text clear...

FLUX.1 Krea [dev] is a text-generated graph model from Black Forest Labs and Krea AI. The model is capable of generating high-quality, photorealistic images based on input text descriptions with a unique aesthetic style that avoids traditional A...

Qwen3-Coder-Flash is a high-performance programming model introduced by Ali Tongyi Thousand Questions team, which has excellent agent-based programming and tool invocation capabilities, and is good at handling complex programming tasks. The model supports 256K tokens of long context understanding, and can scale to 1M ...

Wide Research is a powerful feature of the Manus platform designed to handle complex and large-scale tasks. The platform supports hundreds of general-purpose intelligences working simultaneously through system-level parallel processing mechanisms and intelligence collaboration protocols.

Seed Diffusion is an experimental diffusion language model introduced by ByteHop that handles code generation tasks. The model is based on techniques such as two-stage diffusion training, constrained sequential learning, and enhanced efficient parallel decoding, which significantly improves inference speed to 2146 tokens/s, which is faster than...

Xiao Xingxu is the AI emotion comic generation product launched by Jingdong Health, currently in the testing stage, the product is emotionally driven comic story generation as the core function, the user can express emotion or tell a story through voice or text input, the AI generates matching four-panel comics and story interpretation based on the input.

1688 AI version is an intelligent business assistant application launched by Alibaba's 1688 platform, designed for small B buyers and merchants. Based on the massive data of 1688 platform, the application provides business opportunity push, product recommendation, idea generation, enterprise inquiry and other functions to help users accurately grasp the market dynamics, rapid...

Steps Deep Research is an efficient AI research tool launched by Steps Star, which can autonomously complete research on complex issues and generate professional reports in a short period of time. The tool is designed for finance, consulting, healthcare, law and other fields, and excels in industry reviews with its in-depth search and information integration capabilities.

Runway Aleph is an advanced AI video editing model launched by Runway, which is based on simple text commands to quickly add and delete video content, style change, environment adjustment and camera movement optimization. Users can easily remove redundant elements, change scenes without complex operations...

WebShaper is an AI training data synthesis system launched by Alibaba's Tongyi Lab, which is based on formal modeling and intelligence expansion mechanism to generate high-quality and scalable training data to help AI intelligences improve complex information retrieval capabilities. The system introduces the concept of "knowledge projection"...

Skywork UniPic is an open source multimodal pre-training model from KunlunWanwei, with three core capabilities: image understanding, text generation image and image editing. The model is based on an autoregressive architecture, incorporating a MAR encoder and a SigLIP2 backbone, with 1.5B parameter gauge...

ChatGPT Study is an innovative learning model from OpenAI that helps users learn and understand more efficiently.ChatGPT Study guides users to think actively and solve problems step-by-step through Socratic questioning, scaffolded responses, and personalized instruction....

Tongyi Wanphase Wan2.2 is an advanced AI video generation model open-sourced by Alibaba, with 27 billion total references. The model contains three modes of text-generated video, graph-generated video, and unified video generation, which can generate high-quality videos based on text descriptions, images, or a combination of both.

GLM-4.5 is an open source SOTA model from Smart Spectrum, designed for intelligent body applications, incorporating reasoning, code generation and intelligent body capabilities. The model is based on the Mixed Expert (MoE) architecture and contains two versions, GLM-4.5 with 355 billion parameters and 106 billion...

Coze Studio is ByteDance's open source AI Intelligent Body development platform designed for developers to simplify the building, deployment and management of AI applications.Coze Studio provides a one-stop development environment that supports Prompt, RAG, Plugin...

Coze Loop is the open source AI intelligence development and operation and maintenance management platform of Coze Platform under ByteDance. The platform provides developers with full lifecycle management from development, debugging to evaluation and monitoring, covering cue word engineering, Agent effect evaluation, performance monitoring and tuning...