FLUX.2 is a visual intelligence model from Black Forest Labs, designed specifically for real-world creative workflows. The model supports multi-image references of up to 10 images, generating high-quality images up to 4MP resolution, and boasts extremely high...
ViMax is an end-to-end multi-agent video generation framework developed by the Data Science Lab at the University of Hong Kong. It supports the automatic conversion of ideas, scripts, or novels into complete videos. The framework integrates the roles of director, screenwriter, producer, and video generator...
Z-Image is an image generation model launched by Alibaba Tongyi, featuring 6B parameters. The model includes three variants: Z-Image-Turbo, Z-Image-Base, and Z-Image-Edit, which excel in fast inference, basic development, and image editing, respectively...
Depth Anything 3 (DA3) is a visual space reconstruction model launched by ByteDance's Seed team. It uses a single Transformer architecture to recover the three-dimensional spatial geometry from visual input from any viewpoint.
DeepSeek-Math-V2 is an open-source mathematical reasoning model from the DeepSeek team that enables self-verifying mathematical reasoning. The model focuses on the correctness of the answer and emphasizes the rigor of the reasoning process.
StepAudio R1 is the world's first open-source native audio inference model, developed by the StepAudio team. The model addresses the performance degradation of traditional audio models in complex inference scenarios through its innovative Modal Anchored Inference Distillation (MGRD) framework...
Ito is an open-source intelligent speech dictation tool that uses its unique "VibeTyping" technology to convert users' spoken intentions into beautifully formatted text. It supports various platforms including email clients, document processing, chat platforms, browsers, and code editors...
GELab-Zero is an open-source GUI Agent model developed by StepFun, focusing on automated interaction and task execution on mobile devices. It features local deployability, supports running the 4B model on consumer-grade hardware, and ensures low latency...
Vidi2 is a multimodal large language model launched by ByteDance, focusing on video understanding and creation. It has achieved industry-leading performance in multimodal temporal retrieval (TR), and also excels in spatiotemporal localization (STG) and video question answering (Video QA)...
Keling O1 (Keling Video O1 Model) is the world's first unified multimodal video generation model launched by Keling AI. The model achieves seamless integration of video generation, editing and understanding through an innovative multimodal visual language (MVL) architecture.
Gen-4.5 is a video generation model developed by RunWay. This model sets new industry standards in motion quality, visual realism, and cue word adherence in video generation. Gen-4.5 can generate cinematic, extremely realistic footage...
PixVerse V5.5 is a large-scale video generation model launched by PixVerse Technology. Through its self-developed multimodal understanding and generation technology, the model supports simultaneous audio and video generation, simplifying the complex workflow from conception to final product and improving creative efficiency.
Alpamayo-R1 (AR1) is a Visual-Language-Action (VLA) model from NVIDIA that improves the decision-making capabilities and generalization of autonomous driving through causal reasoning. The model's core innovations include: constructing a CoC (CoC) dataset...
NEO is a novel multimodal model architecture jointly developed by SenseTime and Nanyang Technological University. As the first native multimodal architecture (Native VLM), NEO breaks through the limitations of traditional multimodal models through deep innovation in its underlying architecture...
Mistral 3 is a new generation of open-source AI models from Mistral AI, including the small Ministral 3 (3B, 8B, 14B parameters) and the large Mistral Large 3 (675B total parameters, 41B activation parameters...).
Amazon Nova 2 is a suite of advanced AI models from Amazon Web Services (AWS), designed to meet the diverse needs of enterprises. The Amazon Nova 2 family includes four models: Nova 2 Lite (cost-optimized text generation...
MemMachine is an open-source AI memory system designed specifically for AI agents, providing short-term, long-term, and personalized memory capabilities. MemMachine helps AI applications learn and store information through episodic memory, semantic memory, and user profile memory...
TalkCody is an open-source AI programming assistant that helps developers build software more efficiently. Built with Rust and Tauri, it boasts native performance and low resource consumption. TalkCody supports seamless switching between multiple models and...
Seedream 4.5 (Doubao-Seedream-4.5) is an image creation model developed by ByteDance, specifically designed for commercial productivity scenarios. The model excels in subject consistency, instruction accuracy, spatial logic, and aesthetic expressiveness...
Keling 2.6 is an innovative AI video creation model launched by the Keling AI team. It achieves synchronous audio and video generation and can automatically generate videos containing natural speech, matching sound effects and environmental atmosphere through text or image input.
Flowra is an open-source AI workflow development tool developed by ModelScope in collaboration with the Wuli team. It is the core engine of FlowBench. The tool provides a graph execution engine and a node package development toolchain, supporting a rich type system, ...
Gemini 3 Deep Think is a powerful reasoning model launched by Google DeepMind. It has the ability to think in parallel, explore multiple hypotheses at the same time, and select the optimal answer path.
Workspace Studio is an AI automation tool launched by Google. Users can describe tasks in natural language and quickly generate customized AI agents based on the Gemini model to automate complex workflows. The tool...
RoboCOIN is an embodied intelligence dataset released by the Beijing Academy of Artificial Intelligence in collaboration with several universities and enterprises. It contains over 180,000 demonstration data points, covering 421 tasks and 16 different scenarios, such as home, business, and factory. ...