LLM 与基础模型
DeepSeek-Coder：Code generation models
deepseek-ai/DeepSeek-Coder
ToolAI 对 deepseek-ai/DeepSeek-Coder 的客观项目介绍。DeepSeek Coder: Let the Code Write Itself
★ 24.2K⑂ 2.9KPython
ToolAI.io · GitHub 频道
GitHub 上的开源 AI 项目
一个基于事实的 LLM、智能体、MCP、RAG 和 AI 开发项目索引；每个条目均包含许可证、设置、下载、截图及相关资源。
精选
浏览列表 →markitdown：Document conversion and extraction
ToolAI 对 microsoft/markitdown 的客观项目介绍。Python tool for converting files and office documents to Markdown
★ 174.9K · Pythonskills：Reusable agent skills and workflows
ToolAI 对 anthropics/skills 的客观项目介绍。Public repository for Agent Skills
★ 170.6K · Markdownclaude-code：Agentic coding in the terminal
ToolAI 对 anthropics/claude-code 的客观项目介绍。Claude Code is an agentic coding tool that lives in your terminal, understands your codebase, and helps you code faster by executing routine tasks, explaining complex code, and handling git workflows - all through natural language commands
★ 142.1K · TypeScriptwhisper：Speech recognition and transcription
ToolAI 对 openai/whisper 的客观项目介绍。Robust Speech Recognition via Large-Scale Weak Supervision
★ 107.7K · PythonDeepSeek-V3：Large language model research and inference
ToolAI 对 deepseek-ai/DeepSeek-V3 的客观项目介绍。An open-source project focused on large language model research and inference.
★ 104.4K · PythonDeepSeek-R1：Reasoning model research and evaluation
ToolAI 对 deepseek-ai/DeepSeek-R1 的客观项目介绍。An open-source project focused on reasoning model research and evaluation.
★ 92K · PythonHermes Agent
由 Nous Research 构建的自改进 AI 智能体，具备闭环学习、多平台消息集成以及灵活的模型提供商切换功能。设计用于在低成本 VPS、GPU 集群或无服务器基础设施上运行。
★ 227.1K · PythonHugging Face Transformers
一个面向文本、视觉、音频和多模态领域最先进机器学习模型的模型定义框架，同时支持推理和训练。
★ 163.3K · PythonFirecrawl
一款开源的网络上下文 API，支持大规模搜索、抓取网页并与网络交互，将网页数据转换为干净的 Markdown 或结构化 JSON，供 AI 智能体和大语言模型（LLM）使用。
★ 161.1K · TypeScript MCP 与工具调用Dify：用于智能体工作流和 RAG 应用的协作平台
Dify 是一个 LLM 应用开发平台，用于可视化构建工作流、RAG 管道和可调用工具的智能体。它在协作工作区中整合了模型管理、提示词开发、可观测性和应用 API，可作为托管服务使用或自行托管。
★ 151.2K · TypeScriptLangChain
一个开源的 Python 框架，通过可互操作、可链式组合的组件和第三方集成，构建由大语言模型驱动的应用与智能体。
★ 143.6K · PythonGraphify：将任意代码库转化为可查询的知识图谱
Graphify 是一款本地优先的 CLI 工具和 AI 助手技能，通过确定性的 AST 解析，将代码库、文档、SQL 架构和 PDF 解析为可查询的知识图谱，无需依赖向量存储或嵌入。
★ 104K · Python公开仓库数据
项目索引
多模态 AI
DeepSeek-OCR：Document OCR and visual understanding
deepseek-ai/DeepSeek-OCR
ToolAI 对 deepseek-ai/DeepSeek-OCR 的客观项目介绍。Contexts Optical Compression
★ 23.8K⑂ 2.2KPython
智能体与多智能体
NemoClaw：Agent runtime and deployment tooling
NVIDIA/NemoClaw
ToolAI 对 NVIDIA/NemoClaw 的客观项目介绍。Run agents like Hermes, LangChain Deep Agents, and OpenClaw more securely inside NVIDIA OpenShell with managed inference
★ 22.2K⑂ 3KPython
LLM 与基础模型
gpt-oss：Open-weight language model tooling
openai/gpt-oss
ToolAI 对 openai/gpt-oss 的客观项目介绍。gpt-oss-120b and gpt-oss-20b are two open-weight language models by OpenAI
★ 20.3K⑂ 2.1KPython
多模态 AI
Qwen3-VL：Vision-language models
QwenLM/Qwen3-VL
ToolAI 对 QwenLM/Qwen3-VL 的客观项目介绍。Qwen3-VL is the multimodal large language model series developed by Qwen team, Alibaba Cloud
★ 19.8K⑂ 1.8KPython
推理、部署与运行时
tiktoken：Fast tokenization
openai/tiktoken
ToolAI 对 openai/tiktoken 的客观项目介绍。tiktoken is a fast BPE tokeniser for use with OpenAI's models
★ 19K⑂ 1.6KRust
多模态 AI
Janus：Unified multimodal understanding and generation
deepseek-ai/Janus
ToolAI 对 deepseek-ai/Janus 的客观项目介绍。Janus-Series: Unified Multimodal Understanding and Generation Models
★ 17.8K⑂ 2.2KPython
智能体与多智能体
Qwen-Agent：Tool-using and RAG agents
QwenLM/Qwen-Agent
ToolAI 对 QwenLM/Qwen-Agent 的客观项目介绍。Agent framework and applications built upon Qwen>=3.0, featuring Function Calling, MCP, Code Interpreter, RAG, Chrome extension, etc
★ 17K⑂ 1.7KPython
智能体与多智能体
SkillOpt：Skill optimization for agents
microsoft/SkillOpt
ToolAI 对 microsoft/SkillOpt 的客观项目介绍。SkillOpt is a text-space optimizer that trains reusable natural-language skills for frozen LLM agents through trajectory-driven edits, validation-gated updates, and deployable best_skill.md artifacts
★ 16.2K⑂ 1.5KPython
AI 编程与开发者工具
kimi-cli
MoonshotAI/kimi-cli
Kimi Code CLI is your next CLI agent.
★ 10.8K⑂ 1.3KPython
智能体与多智能体
kimi-code
MoonshotAI/kimi-code
Kimi Code CLI — The Starting Point for Next-Gen Agents
★ 5.1K⑂ 742TypeScript
推理、部署与运行时
Model-Optimizer
NVIDIA/Model-Optimizer
A unified library of SOTA model optimization techniques like quantization, distillation, pruning, neural architecture search, speculative decoding, etc. It compresses deep learning models for downstream deployment frameworks like TensorRT-LLM, TensorRT, vLLM, etc. to optimize inference speed.
★ 3.3K⑂ 513Python
第 13 / 15 页 · 173 个项目