PhoneBuddy is an open-source 4B parameter mobile agent model developed by Tencent's Hunyuan team. Its core research focuses on how to train a usable AI agent in real-world mobile scenarios.
Mistral OCR 4 is the latest generation document understanding model from Mistral AI. The model supports extracting text from complex documents such as PDFs, images, and presentations, and can return bounding box localization, region type classification, and confidence scores...
SkillOpt is an open-source agent-based skill document optimization tool from Microsoft. The tool introduces deep learning training logic into the text space, treating skill.md as trainable parameters, generating editing directions through trajectory analysis, and constraining them with text learning rates...
Confucius4-TTS is an open-source 1.3B parameter multilingual speech synthesis engine from NetEase Youdao. The model can clone voices with zero samples using only 3 seconds of reference audio, without requiring reference text, and supports cross-language processing of 14 languages including Chinese, English, Japanese, and Korean...
MMSkills is a multimodal skill framework for general-purpose visual agents, jointly launched by Shanghai Jiao Tong University and Xiaohongshu. It supports expanding reusable skills from plain text steps to include text flows, runtime state cards, and multi-view functionality...
Doubao Audio Generation Model 1.0 is an end-to-end audio creation model launched by Volcano Engine, supporting text or audio as reference input to generate target audio. A single Prompt can be edited with multi-character dialogue, emotional tone, background music, and...
Baichuan-M4 is a next-generation medical augmentation model jointly developed by Baichuan Intelligent and Tsinghua University. It simultaneously ranks first in the world on three authoritative health benchmarks: HealthBench Comprehensive, Hard, and Professional, with a hallucination rate as low as 3.3%...
JoyAI-VL-Interaction is an open-source real-time video visual language interaction model from JD.com's Joy Future Academy. It is the world's first full-stack open-source interaction model system.
Qwen-AgentWorld is the first language world model launched by the Tongyi Qianwen team. It simulates seven intelligent agent environments, including MCP, search, terminal, software engineering, Android, Web, and operating system, through long thought chain reasoning.
Seedance 2.5 is the latest flagship version of ByteDance's Doubao video generation model, expected to be fully launched in early July. As a major upgrade to Seedance 2.0, the model achieves three global breakthroughs, including single-segment native video...
Unlimited-OCR is an end-to-end long document OCR model launched by Baidu. It uses a Reference Sliding Window Attention mechanism to compress the decoder's key-value cache from linear growth to a constant, achieving single forward transformation...
AudioX-Turbo is a unified and efficient audio generation framework launched by Noiz AI in collaboration with Hong Kong University of Science and Technology and Tsinghua University. It supports the generation of high-quality sound effects and music by combining any combination of text, video, and audio inputs.
Boogu-Image-0.1 is an open-source family of unified image generation and editing models from the Boogu team. The models support text-to-image generation, imperative image editing, and bilingual (Chinese and English) text rendering within the same architecture. The family includes Base, Edit...
Spatial-TTT is a streaming visual spatial intelligence framework jointly developed by Tsinghua University, Tencent Hunyuan, and Nanyang Technological University. The framework has only 2B parameters and uses Test-Time Training (TTT) technology to train the user during video streaming...
Seed 2.1 is a new series of intelligent agent models launched by ByteDance's Seed team for real-world productivity scenarios, including Pro and Turbo versions.
VidMuse is an AI music and video agent launched by Sand.ai. Users only need to upload a Suno link or MP3 file, and it can automatically generate a 30-60 second music video. VidMuse 2.0 is based on the Video as Code concept,...
HappyHorse 1.1 is the latest upgraded version of Alibaba's AI video generation model, which has achieved a systematic upgrade in five dimensions compared to version 1.0: dynamic expressiveness, subject consistency, instruction compliance, visual quality, and audio capabilities.
Skill Zoo is a one-stop desktop skill management tool for Coding Agents. The tool uses an SSOT+ symbolic link architecture to uniformly manage skills scattered across programming tools such as Claude Code, Codex, Trae CN, and Hermes...
Cowart is an open-source plugin for OpenAI Codex, developed by Zhong Erxin, product manager of Doubao Desktop. It's a local infinite canvas tool built on the tldraw framework, allowing Codex to place images on a local electronic whiteboard...
Xiaomi Miloco 2.0 is an open-source whole-house smart AI solution launched by Xiaomi based on the MiMo-V2.5-Pro big model. It connects via OpenClaw as an Agent, enabling multimodal perception, proactive intelligence, continuous task management, and more...
Sakana Fugu is a multi-agent orchestration system from Sakana AI that dynamically schedules top-performing models using a single API. The system assigns thinker, executor, and verifier roles, automatically handling selection, delegation, and synthesis without requiring pre-defined workflows...
SpatialClaw is a training-free spatial reasoning framework jointly developed by NVIDIA Research and KAIST. It iteratively executes complex 3D/4D spatial reasoning tasks within a persistent Python kernel using a code-as-action mechanism. (The last sentence appears to be incomplete and possibly refers to a separate, unrelated point.)
MaineCoon is the world's first real-time audio and video autoregressive world model optimized for social interaction scenarios. The model boasts 22 billion parameters and can achieve real-time streaming generation at 47.5 FPS on a single GPU, supporting sub-second interactive responses...
LOGOS (Language Of Generative Objects in Science) is the first unified scientific grammar and multi-domain scientific generation foundation open-sourced by Alibaba's ATH-Token Foundry in collaboration with Renmin University of China's Hillhouse Institute of Artificial Intelligence...