`last30days-skill` is an open-source AI agent research skill that allows agents to automatically scrape real discussions from major overseas social media platforms over the past 30 days and synthesize citation-supported reports. The tool deeply crawls Reddit comments, X High...
WorkBuddy Bench is an open-source coding intelligence evaluation suite from Tencent, covering four real-world scenarios: code, web development, office work, and security. Tasks are reverse-engineered from real commits, pull requests, and business scenarios, and rewritten into conversational requests...
MineExplorer, developed by the Meituan LongCat team, is the first benchmark for evaluating minute-level long-duration open-world tasks based on Minecraft. MineExplorer includes 813 manually verified instances with 1-4 jumps, and the tasks contain hidden prerequisites...
The Claude Opus 5 is Anthropic's latest flagship model, positioned as "a cutting-edge smart device close to the Fable 5, priced at about half the latter." Compared to the Opus 4.8, it offers significantly improved performance without changing costs...
BigMac is an open-source multimodal large model pipelined parallel training framework developed by the Dots Infra team at Xiaohongshu. It addresses the challenge of balancing computational efficiency and memory usage in MLLM training by proposing a secure nested pipeline paradigm.
Cosmos 3 Edge is a 4B parameter open-source world model from NVIDIA, designed for real-time inference in robotics and edge AI. The model is based on the Nemotron architecture, incorporating both autoregressive and diffusion-based Transformer towers...
CodeBuddy NPC is a cloud-based AI agent launched by Tencent Cloud for enterprise R&D processes. It directly intervenes in code repositories, issues, PRs, and CI/CD processes by relying on the CNB platform.
Cursor Router is an intelligent model router launched by Cursor for teams and enterprises. It is trained based on more than 600,000 real requests and automatically routes each programming request to the most suitable model.
MAI-Image-2.5-Pro is a high-precision image generation model developed by Microsoft's AI team. The model focuses on high-quality main visual generation, detailed editing, and in-image text rendering, and supports content adjustment via natural language commands.
MAI-Voice-2-Flash is a high-speed speech synthesis model developed by Microsoft's AI team. The model is specifically optimized for high-concurrency, low-latency speech scenarios, offering approximately twice the speed and 32% lower cost compared to its predecessor, MAI-Voice-2, while maintaining...
Ling-3.0-Flash is a lightweight MoE inference model launched by Ant Group's Bailing Big Model. It has a total of 124B parameters, with 5.1B parameters activated per token. It supports 256K ultra-long context and thinking/non-thinking dual modes.
img2threejs is an open-source AI skill that automatically converts a single reference image into an interactive Three.js 3D model. Users simply provide an image, and the AI analyzes the object's outline, proportions, material, and details to generate...
SciReasoner is a next-generation multimodal scientific reasoning model developed by the Shanghai Artificial Intelligence Laboratory in collaboration with teams from the Chinese University of Hong Kong, the Institute of Materia Medica of the Chinese Academy of Sciences, Stanford University, Oxford University, and others. The model is the first to propose a native structured reasoning paradigm, integrating egg...
OpenAI Presence is an enterprise-grade AI agent deployment and management platform launched by OpenAI, which helps enterprises advance intelligent agents from the experimental stage to the production environment.
SenseNova U1 Pro is SenseTime's flagship image creation model, with a core focus on natively unifying understanding, generation, and action. The model supports native 8K output, possesses the ability to integrate text and images, and can continuously complete tasks around a target...
Macaron V1 is a post-trained multi-capability model system from MindLab. Based on the GLM 5.2 platform, it uses reinforcement learning to fine-tune approximately 0.5% of the core parameters using LoRA. The system features a unique Mixture-of-LoRA (MoL) method...
UnifoLM-OminiA-0.3 is a multimodal interactive understanding model launched by Unitree Robotics, designed specifically for the G1 humanoid robot. A single model can coordinate multiple tasks in home and elderly care settings, supporting visual recognition, voice interaction, fine motor skills, and device control...
Gemini 3.6 Flash is Google DeepMind's latest flagship working model, designed for large-scale AI agent scenarios. It includes three new Gemini models: Gemini 3.6 Flash, 3.5 Flash-Lite, and 3...
The Gemini 3.5 Flash-Lite is Google's fastest and cheapest Gemini 3.5 series model, designed for high-throughput, low-latency production-grade agent workflows.
Gemini 3.5 Flash Cyber is a dedicated cybersecurity model fine-tuned by Google based on Gemini 3.5 Flash. It is used in conjunction with the CodeMender code security agent and specializes in vulnerability discovery and remediation.
Qwen-Image-3.0 is the third-generation image generation model launched by the Alibaba Tongyi Qianwen team. The model supports ultra-long inputs of up to 4.5k tokens and can render complex nine-grid knowledge graphs in one go; it also supports precise typesetting of 10px small text...
Audio-Visual Flamingo (AV-Flamingo for short) is an open-source audio-visual large language model jointly developed by NVIDIA and the University of Maryland. It is specifically designed for the joint understanding and inference of long videos and complex real-world audio-visual content...
Hyra (Hunyuan Research Agent) is a recursive, self-improving research agent developed by Tencent Hunyuan. The first version, Hyra-1.0, has been officially released. Hyra uses the lightweight Harness framework and utilizes a Context Agent...
AskXiaoBai 5 Pro is a new generation long content generation engine launched by YuanShi Technology. It is built on the self-developed YuanShi large model and the full version of DeepSeek-R1 671B, and focuses on three major capabilities: long content, stability and searchability.