Meituan GEO Marketing Portal is a Generative Engine Optimization (GEO) platform launched by Meituan, designed specifically for the AI search era. It helps brands increase exposure and recommendation frequency across major search engines such as Doubao, DeepSeek, and Wenxin Yiyan. ...
SenseNova-U1-8B-MoT-Infographic is an open-source 8B parametric information graph enhancement model from SenseTime. Based on the unified architecture of SenseNova-U1-8B-MoT, it significantly improves performance through specialized data training and reinforcement learning (RL)...
PilotDeck is an open-source agent operating system jointly developed by Tsinghua University's THUNLP Lab, Wallfacer Intelligence, OpenBMB, and AI9stars. PilotDeck uses a WorkSpace-based design paradigm, creating independent files for each project...
Qwen-Image-Bench is a dataset for evaluating text-based image models, developed by the Tongyi Qianwen team. It contains 1,000 test samples, provides both Chinese and English prompts, and supports multi-dimensional evaluation of the generation performance of various text-based image models. It utilizes APIs...
Claude Opus 4.8 is Anthropic's flagship large language model. Building upon Opus 4.7, it improves judgment, honesty, and long-term independent operation capabilities, performing well in benchmark tests such as programming, agent reasoning, and multidisciplinary reasoning...
Step 3.7 Flash is a new generation of open-source Flash model launched by StepStar for the Agent production stage. It adopts a sparse MoE architecture and has a maximum generation speed of 400 Tokens/s.
Hy-Memory is a memory plugin developed by Tencent Hunyuan, specifically designed for long-term collaborative agents such as OpenClaw. Through a 6-layer memory framework × System1/System2 dual-system × evolution chain three-layer architecture, it enables agents to truly...
Dynamic Workflows, introduced by Claude Code, is currently in research preview. It supports dynamically writing orchestration scripts in Claude, allowing for the parallel scheduling of dozens to hundreds of sub-agents within a single session...
Alibaba Cloud's Bailian CLI is an open-source AI Agent command-line tool designed specifically for intelligent agent scenarios. With a single command, developers can enable the agent to automatically access over 150 multimodal models, 10+ applications, and knowledge bases from the Alibaba Cloud Bailian platform...
Fara 1.5 is the latest browser-based AI agent (CUA) model series released by Microsoft Research AI Frontiers Lab, including three parameter versions: 4B, 9B, and 27B. The model is based on Qwen 3.5 with minor tweaks and uses pure pixel-based...
Guizang Social Card Skill is an open-source image and text card technology adapted to Agent environments such as Claude Code/Codex. It can solve the problem of adding images for Xiaohongshu (3:4 vertical images) and WeChat Official Accounts (21:9 + 1:1 cover images)...
Wall-OSS-0.5 is an open-source, domestically developed embodied intelligence model for independent variables robots. It belongs to the large-scale Vision-Language-Action (VLA) model category and possesses the 'zero-shot' capability to be deployed on real robots without task fine-tuning. The model innovates...
MAI-Image-2.5 is Microsoft Research's flagship raw image model, the most powerful version in the MAI-Image series. The model achieved a score of 1,254 on the Arena raw image leaderboard, ranking 3rd, a 72-point improvement over its predecessor...
omp (oh-my-pi) is an open-source AI terminal programming agent based on the Pi project. Its core code consists of approximately 27,000 lines of Rust and supports 40+ model vendors and 32+ built-in tools.
ForgeTrain is the world's first production-grade large model pre-training framework, jointly released by Facewall Intelligence, Tsinghua University, and the OpenBMB open-source community. It is entirely written by AI and requires zero human code intervention.
opera-browser-cli is an open-source command-line tool from Opera Neon, built on top of opera-devtools, that supports local AI agents (such as Claude Code) to directly control the browser via the terminal.
Rodin Gen-2.5, developed by Hyper3D, is the world's first AI 3D model supporting the generation of tens of millions of polygons. Based on the best paper technology from SIGGRAPH 2025, it supports text-based 3D, image-based 3D, and multi-viewpoint 3D...
Aholo Viewer is an open-source, high-performance web-based rendering engine developed by Quncore Technology, based on 3D Gaussian sputtering (3DGS). The tool employs chunk-level LOD streaming technology, allowing 3D content to be rendered like video on different devices...
Keye-VL-2.0-30B-A3B is a self-developed multimodal large model open-sourced by Kuaishou, serving as the main foundation for 30B-level models. This model is the first to introduce DSA sparse attention into multimodal scenarios, supporting ultra-long contexts of up to 256K, enabling hour-level video...
Xiaomi Auto World Model is an assisted driving world model launched by Xiaomi Auto, which is the first to deeply couple 3D reconstruction (WorldRec) with video generation (WorldGen).
SkyClaw-v1.0 is a high-performance agent model launched by Kunlun Tech AI. It supports long contexts with millions of tokens and is deeply adapted to mainstream agent frameworks such as OpenClaw, Hermes, Claude Code, and Codex.
AgentScope 2.0 is an open-source multi-agent development framework launched by Tongyi Labs. It focuses on stable operation, security control, and production deployment in real-world scenarios, adding features such as model fault tolerance, a permission system, refactored context management, and Middleware...
MiniCPM5-1B is a 1B parameter edge-side text-based large model jointly developed by Wallfacer Intelligence, Tsinghua University, and the OpenBMB open-source community. It achieved a score of 17.9 on the AA-Index leaderboard, surpassing all models below 2B and becoming the world's leading model in its class...
Dulus is an open-source CLI AI agent consisting of approximately 12,000 lines of Python code, supporting over 40 models including Claude, GPT, Gemini, DeepSeek, Kimi, and Qwen. The tool can directly capture visitor sessions for the web-based AI application in a browser...