Clawith is an open-source hybrid multi-agent collaboration framework for enterprises, treating AI agents as 'digital employees' rather than simple chat tools. Each agent possesses a persistent identity (soul.md), long-term memory (memory.md), and independent...
Kairos 3.0-4B is an open-source embodied native world model from DaXiao Robotics, pioneering an integrated architecture of 'multimodal understanding-generation-prediction'. As the world's first lightweight 4B model capable of end-side robot body control, it is implemented on the THOR platform...
InternVL-U is a lightweight, unified multimodal model with 4B parameters, open-sourced by the Shanghai Artificial Intelligence Laboratory in collaboration with several top universities. It achieves an end-to-end closed loop of 'understanding-reasoning-generation-editing' for the first time. The model employs 'unified contextual modeling + modal...'
Solaris is the first multiplayer video world generation model that can simultaneously generate a consistent first-person perspective for two players in Minecraft. It overcomes the limitation of existing models that only support single-player mode, ensuring spatial consistency across player perspectives...
StepClaw is a cloud-based intelligent assistant launched by StepAI, based on the OpenClaw architecture. Users can quickly deploy it in the cloud through the StepAI APP, saving them the trouble of local hardware configuration.
LTX-2.3 is the latest generation video generation model open-sourced by the Israeli AI company Lightricks. It employs the Diffusion Transformer architecture and boasts 22 billion parameters. The model supports text, image, and audio input...
Nemotron 3 Super is an open-source AI model from NVIDIA with 120 billion parameters. It employs a Mamba-MoE hybrid architecture and is optimized for agent applications. The model supports ultra-long contexts with up to 1 million tokens, offering a 3x speedup inference and high throughput...
Gemini Embedding 2 is Google's first native multimodal embedding model, built on the Gemini architecture. The model maps text, images, video, audio, and documents to a unified vector space, supporting over 100 languages...
CLI-Anything is an open-source tool from the Data Science Laboratory (HKUDS) at the University of Hong Kong that can convert the codebase of any open-source software into a command-line interface (CLI) usable by AI Agents with a single click.
SongGeneration 2 is a 4B-parameter music generation model jointly open-sourced by Tencent and Tsinghua University. It employs a hybrid LLM-diffusion architecture and hierarchical representation design to achieve commercial-grade music generation quality. The model achieves a lyrics accuracy of 8.55%...
MiroFish is an AI prediction engine based on multi-agent technology, developed by BaiFu, a 20-year-old student at the University of Science and Technology of China. MiroFish can automatically construct high-fidelity parallel digital worlds from real-world seed information (news, novels, policies, etc.), generating...
Mobile-Agent-v3.5 is a new generation of multi-platform GUI agent framework open-sourced by Alibaba Tongyi Labs, marking the transition of open-source GUI agents from 'demonstration level' to 'engineering-ready level'. The framework natively supports desktop, mobile, browser...
gws (Google Workspace CLI) is an open-source command-line tool developed by the Google Workspace team. Written in Rust and distributed via npm, its biggest feature is its ability to dynamically build commands at runtime, independent of a static command list...
HY-WU (Hun Yuan Wu Xiang) is a functional neural memory framework launched by Tencent Hunyuan, enabling AI to generate personalized parameters in real time during inference. The framework uses a Transformer parameter generator to synthesize personalized parameters in real time based on the input during inference...
JadeAI is an open-source AI-powered resume generator developed by twwch, positioned as a truly user-friendly, completely free, and self-hostable resume solution. Built on Next.js and released under the Apache 2.0 license, it...
Symphony is an open-source agent orchestration system from OpenAI that supports transforming project work into isolated, autonomous tasks. Symphony can monitor Kanban boards such as Linear and automatically schedule agents to complete coding, testing, merging, and more...
Roubao is an open-source AI-powered mobile automation assistant, intended as a free alternative to ByteDance's 'Doubao Mobile Assistant'. Roubao is based on a Visual Language Model (VLM) and developed natively in Kotlin, requiring no computer or root access...
HiClaw is an open-source agent-based team collaboration system from Alibaba, positioned as a 'Team version of OpenClaw'. The system introduces a Manager Agent as an AI steward to automatically coordinate multiple Worker Agents to complete complex tasks.
Yuan3.0-Ultra is an open-source, trillion-parameter multimodal foundational model from Inspur Information's YuanLab.ai team, with a total of 1.01T parameters and 68.8B activation parameters. It adopts a MoE architecture, integrating a visual encoder and a multimodal alignment module...
Helios is a 14B-parameter real-time long video generation model jointly developed by Peking University and ByteDance, capable of generating minute-level videos at 19.5 FPS on a single H100 GPU.
Phi-4-reasoning-vision-15B is an open-source 15 billion parameter multimodal reasoning model from Microsoft, built on the Phi-4-Reasoning language model and the SigLIP-2 visual encoder.
MiniMax Music 2.5+ is an AI music generation model launched by MiniMax, focusing on pure music creation capabilities. The model supports diverse styles such as classical orchestral, electronic, and ambient sounds, covering all scenarios including meditation, sleep aid, gaming, and film.
MagicAgent is a basic intelligent agent model jointly launched by Honor and Fudan University. It achieves the strongest performance within 10 billion parameters with a 32B dense architecture and a 30B-A3B MoE architecture.
Gemini 3.1 Flash-Lite is Google's lightweight flagship model, emphasizing extreme cost-effectiveness. With an output speed of 363 tokens per second and an input price of $0.25 per million tokens, it significantly outperforms GPT-5 in speed...