Nemotron-Labs-Diffusion is a three-mode language model from NVIDIA that unifies autoregressive, diffusion, and self-inferential decoding within a single architecture. Through joint AR-diffusion target training, the model can operate in various concurrent scenarios...
MoWorld is the world's first high frame rate interactive world model launched by Mochip Technology. It is based on the fully domestic Huawei Ascend NPU full-stack optimization, achieving a maximum of 50 FPS real-time inference, and reducing the cost by 70% compared to GPU.
LingBot-Vision is an open-source general-purpose visual foundation model from Ant Lingbo Technology. It is the first in the industry to use boundary structures as pre-training targets and adopt geometric modeling to achieve a breakthrough in spatial perception training paradigm.
motion-anything is an open-source, native-first motion engine from the Open Design team, positioned as a free version of Figma Motion + a watermark-free motion video workshop. The tool allows motion effects to come alive directly on real web pages, as you see...
Muse Image is the first self-developed AI image generation model launched by Meta Super Intelligence Lab. It adopts an Agent architecture and collaborates with Muse Spark for multi-step planning and self-refinement.
SayIt is an open-source AI voice input method built on Rust and supports Windows desktop. Users can press and hold a shortcut key to speak, and their spoken words will be converted into usable written text in real time.
LingBot-Depth 2.0 is a spatial perception model for real-world scenarios launched by Antminer Technology, along with the open-source visual foundation model LingBot-Vision. The training data has been expanded from 3 million to 150 million, and includes 16 deep learning algorithms...
OpenScience is an open-source AI research platform launched by Synthetic Sciences, targeting research in machine learning, biology, physics, and chemistry. The platform supports processes from literature review, hypothesis generation, code writing, and experiment execution...
GenEvolve is a self-evolving image generation agent jointly developed by the Hong Kong University of Science and Technology (Guangzhou), Meituan, and the National University of Singapore. GenEvolve formalizes open-ended image generation as a visual trajectory orchestrated by tools. The Agent...
GPT-Realtime-2.1 is a next-generation real-time speech model from OpenAI. The 2.1 flagship version significantly improves alphanumeric recognition, silence/noise handling, and interrupt response capabilities, supports voice, text, and image input, and boasts powerful push functionality...
MemGUI-Agent is a long-range mobile GUI intelligent agent jointly launched by Zhejiang University and Kuaishou. It is designed for mobile automation tasks that are cross-app, multi-step, and long-chain.
InternAgentS is an open-source, domestically developed research intelligent agent workbench from the Shanghai AI Lab, designed for AI for Science scenarios. It integrates disparate processes such as paper reading, experimental analysis, code iteration, remote computing, and research writing into a single platform...
Fun-ASR-Realtime is a large-scale streaming real-time speech recognition model launched by Alibaba's Qianwen platform. The model uses the WebSocket streaming protocol to achieve simultaneous speech and word output, with a latency of hundreds of milliseconds for the first word and extremely low latency for the last word.
Hy3 is a 295-parameter hybrid expert (MoE) model open-sourced by Tencent's Hunyuan platform. The model demonstrates significant improvements in agent capabilities, inference, and long-context tasks, achieving benchmark performance comparable to flagship models with 2-5 times the parameter size on multiple benchmarks.
FuckClaude is a lightweight, SEO-friendly, bilingual (English and Chinese) single-page application (SPA) detection tool used to scan a user's browser environment and determine if they will be flagged as a 'Chinese user' by Claude Code. After clicking run, the tool...
ElementsClaw is the industry's first AI agent for superconducting material discovery, jointly launched by Alibaba DAMO Academy, Renmin University of China, and the University of Chinese Academy of Sciences. It employs a specialized-to-general fusion architecture, integrating the 1 billion-parameter atomic fundamental model Elements...
ComAct (COM-as-Action) is a research paradigm that redefines the automation of professional software operations, proposed by Jiaxin Ai et al. This paradigm recognizes Windows' Component Object Model (COM) as a unified executable...
Page Agent is an open-source, pure front-end JavaScript intelligent agent library from Alibaba. It can be embedded in a webpage with a single line of code, allowing users to manipulate page DOM elements using natural language.
EdgeBench is a benchmarking framework launched by ByteDance's Seed team to evaluate the long-term learning capabilities of autonomous AI agents in real-world environments.
Vidu S1, launched by Shengshu Technology, is a world-leading real-time interactive video foundation model, marking the transition of AI video from offline generation to the era of real-time two-way interaction. Based on an autoregressive diffusion architecture, it supports 540P resolution and 25FPS (maximum...
AReaL 2.0 is an open-source agent-based online reinforcement learning infrastructure developed by Ant Financial in collaboration with Hong Kong University of Science and Technology and Tsinghua University. It enables deployed agents to continuously evolve from real-world interaction trajectories.
Ego Lite is a Chromium-based browser that enables parallel collaboration between humans and AI agents. Ego Lite serves as the infrastructure providing the underlying browser operation capabilities for AI agents.
Video-use is an open-source AI video editing agent from the Browser Use team that replaces traditional timeline operations with natural language dialogue. Users simply place the raw footage into a folder and describe their needs, and the agent automatically completes the editing...
ViiTorVoice, launched by Cloudwise, is the world's first AI speech synthesis model supporting partial editing, topping the Seed-TTS authoritative evaluation list. It employs a NAR non-autoregressive architecture, enabling speech editing as easy as editing a Word document...