Thinker is an open-source embodied intelligent visual language model developed by UBTECH, specifically designed for robotic applications. Its 4B parameters achieved the highest global ranking in nine authoritative benchmark tests.
What is Happy? Happy is an open-source tool that allows users to remotely control Claude Code or Codex running on their computers via mobile phones or web clients. It supports real-time code progress monitoring, voice interaction, message push notifications, and more...
Step 3.5 Flash is the latest open-source foundation model from StepStar, specifically designed for Agent scenarios. The model adopts a sparse MoE architecture with a total of 196 billion parameters, with each token activating only 11 billion parameters, balancing performance and efficiency.
Vidu Q3 is the world's first 16-second AI video model with synchronized audio and video, launched by Shengshu Technology. It's specifically designed for narrative scenarios such as short dramas, comics, and commercials. A single prompt is all it takes to produce a 16-second 1080p video, perfectly capturing the visuals, dialogue, and environment...
UnifoLM-VLA-0 is a general-purpose vision-language-action (VLA) large model open-sourced by Unitree Robotics, based on the Qwen2.5-VL-7B architecture for continuous pre-training.
TTT-Discover (Test-Time Training to Discover) is an AI scientific discovery methodology developed by institutions such as Stanford and NVIDIA. TTT-Discover trains the model using reinforcement learning during the testing phase, with non-frozen weights...
Project Genie is an experimental AI world model prototype launched by Google DeepMind, powered by Genie 3, Nano Banana Pro, and Gemini technologies. Users can create virtual worlds using text prompts or uploaded images...
LingBot-World is an open-source interactive world model from Ant Lingbo Technology. The model learns physical laws and causal relationships from large-scale game environments through a scalable data engine, enabling precise action-driven generation.
Qwen3-ASR is a series of open-source speech recognition models from Alibaba Cloud's Tongyi Qianwen team. It includes two ASR models: a 1.7B high-precision version and a 0.6B high-efficiency version, as well as a dedicated Qwen3-ForcedAligner-0.6B forced alignment model.
SenseNova-MARS is an open-source multimodal autonomous reasoning model from SenseTime, offering two parameter scales: 8B and 32B. As the first Agentic VLM to support deep fusion of dynamic visual reasoning and image/text search, the model can function like an intelligent agent...
MOVA (MOSS Video and Audio) is China's first high-performance open-source end-to-end audio and video generation model, jointly launched by the OpenMOSS team at Shanghai Innovation Academy and MOSI. The model breaks through the limitations of traditional video 'mute'...
SkyReels-V3 is an open-source multimodal video generation model from Kunlun Wanwei, enabling professional-grade video creation with a single architecture. The model can convert still images into dynamic footage, supports intelligent video length extension and cinematic transitions, allowing for...
LongCat-Flash-Lite is a new generation of high-efficiency large language model launched by Meituan. The model adopts an innovative hybrid expert (MoE) + N-gram embedding architecture, with a total of 68.5 billion parameters, activating only about 2.9-4.5 billion parameters per inference step...
MiniMax-M2-her is a role-playing model launched by MiniMax for in-depth AI companionship scenarios, serving as the underlying model for Hoshino/Talkie. The model utilizes three core capabilities: a unique world experience, a rhythmic story progression...
MiniMax Music 2.5 is MiniMax's next-generation AI music creation model, achieving breakthroughs in two major technical challenges: "strong control at the paragraph level" and "physical-level high fidelity." The model supports 14 music structure tags (such as intros...).
Mureka V8 is an AI music big model launched by Kunlun Wanwei. Based on the MusiCoT (Music Chain-of-Thought) technology architecture, it achieves a leap from sound splicing to human-like creative logic.
DeepSpeed-MII is an open-source Python library from the DeepSpeed team that provides efficient model inference. DeepSpeed-MII significantly improves inference performance using innovative techniques such as blocking key-value caching, sequential batch processing, and dynamic SplitFuse...
Lucy 2.0 is a real-time world transformation model launched by Decart AI, enabling high-fidelity video editing to leap from offline rendering to an interactive real-time experience.
NVIDIA Earth-2 is NVIDIA's first fully open-source AI weather forecasting model family, based on three core architectures: Atlas, StormScope, and HealDA, achieving 15-day global medium-range forecasts, 0-6...
Kimi K2.5 is the latest generation of all-around flagship model from Dark Side of the Moon, trained on approximately 15TB of visual and text data. It boasts top-tier code generation and visual understanding capabilities, supports autonomous agent cluster collaboration,...
DeepSeek-OCR 2 is the second-generation OCR model released by the DeepSeek team. By introducing the DeepEncoder V2 architecture, it achieves a paradigm shift from fixed scanning to semantic reasoning. The model employs causal stream querying and dual-stream injection...
LingBot-Depth is an open-source high-precision spatial perception model from Ant Lingbo Technology, designed to solve the challenge of depth perception for robots in transparent, reflective objects (such as glass).
The Vidu Q2 Reference Pro is the world's first "everything can be referenced" video model, which provides creators with powerful features by supporting multimodal input (2 videos and 4 images).
White Tiger-VTouch is the world's first and largest cross-body visual-tactile multimodal dataset, open-sourced by the National and Local Jointly Constructed Humanoid Robotics Innovation Center in collaboration with VTouch Robotics. The dataset includes visual-tactile sensor data, RGB-D data, and more...