Open Source
Open models, tools and ecosystems around AI – why they shape the field and where they beat proprietary systems.
The most important open-source AI projects
Ranked by stars on GitHub – the simplest measure of how many developers are following a project.
Data from GitHub, as of: Aug 29, 2026
Models & Inference
| Rank | Project | What it does | Stars | License |
|---|---|---|---|---|
| 1 | ollama/ollama | Get up and running with Kimi-K2.6, GLM-5.2, MiniMax, DeepSeek, gpt-oss, Qwen, Gemma and other models. | 179,732 +396 this week | MIT |
| 2 | huggingface/transformers | 🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training. | 164,607 +209 this week | Apache-2.0 |
| 3 | ggml-org/llama.cpp | LLM inference in C/C++ | 126,222 +795 this week | MIT |
| 4 | deepseek-ai/DeepSeek-V3 | 104,436 +31 this week | MIT | |
| 5 | vllm-project/vllm | A high-throughput and memory-efficient inference and serving engine for LLMs | 90,394 +511 this week | Apache-2.0 |
| 6 | unslothai/unsloth | Local UI to run and train LLMs and diffusion models. Supports GGUF, MLX, Qwen3.8, Kimi K3, MiniMax-H3, Gemma 4, FLUX and more. | 75,159 +554 this week | Apache-2.0 |
| 7 | sgl-project/sglang | SGLang is a high-performance serving framework for large language models and multimodal models. | 32,713 +345 this week | Apache-2.0 |
| 8 | ml-explore/mlx | MLX: An array framework for Apple silicon | 28,212 +83 this week | MIT |
| 9 | QwenLM/Qwen3 | Qwen3 is the large language model series developed by Qwen team, Alibaba Cloud. | 27,569 +15 this week | not stated |
| 10 | huggingface/peft | 🤗 PEFT: State-of-the-art Parameter-Efficient Fine-Tuning. | 21,606 +16 this week | Apache-2.0 |
Agents & Frameworks
| Rank | Project | What it does | Stars | License |
|---|---|---|---|---|
| 1 | n8n-io/n8n | Fair-code workflow automation platform with native AI capabilities. Combine visual building with custom code, self-host or cloud, 400+ integrations. | 202,798 +514 this week | not stated |
| 2 | langgenius/dify | Build Agentic workflows, RAG pipelines, with rich AI model and tool support on one collaborative workspace. Deploy on cloud, VPC, or self-hosted, so teams move from prototype to production without rebuilding the stack. | 153,834 +451 this week | not stated |
| 3 | langchain-ai/langchain | The agent engineering platform. | 145,245 +340 this week | MIT |
| 4 | browser-use/browser-use | 🌐 Make websites accessible for AI agents. Automate tasks online with ease. | 111,650 +1,309 this week | MIT |
| 5 | modelcontextprotocol/servers | Model Context Protocol Servers | 89,956 +123 this week | not stated |
| 6 | OpenHands/OpenHands | 🙌 OpenHands: AI-Driven Development | 85,561 +596 this week | MIT |
| 7 | microsoft/autogen | A programming framework for agentic AI | 60,692 +82 this week | CC-BY-4.0 |
| 8 | crewAIInc/crewAI | Framework for orchestrating role-playing, autonomous AI agents. By fostering collaborative intelligence, CrewAI empowers agents to work together seamlessly, tackling complex tasks. | 57,791 +232 this week | MIT |
| 9 | langchain-ai/langgraph | Build resilient agents. | 40,674 +309 this week | MIT |
| 10 | openai/openai-agents-python | A lightweight, powerful framework for multi-agent workflows | 29,061 +133 this week | MIT |
| 11 | modelcontextprotocol/python-sdk | The official Python SDK for Model Context Protocol servers and clients | 24,152 +44 this week | MIT |
| 12 | anthropics/claude-agent-sdk-python | 7,999 +33 this week | MIT |
RAG & Data
| Rank | Project | What it does | Stars | License |
|---|---|---|---|---|
| 1 | infiniflow/ragflow | RAGFlow is a leading open-source Retrieval-Augmented Generation (RAG) engine that fuses cutting-edge RAG with Agent capabilities to create a superior context layer for LLMs | 89,588 +431 this week | Apache-2.0 |
| 2 | docling-project/docling | Get your documents ready for gen AI | 65,721 +235 this week | MIT |
| 3 | run-llama/llama_index | LlamaIndex is the leading document agent and OCR platform | 51,912 +67 this week | MIT |
| 4 | milvus-io/milvus | Milvus is a high-performance, cloud-native vector database built for scalable vector ANN search | 45,874 +103 this week | Apache-2.0 |
| 5 | facebookresearch/faiss | A library for efficient similarity search and clustering of dense vectors. | 40,819 +18 this week | MIT |
| 6 | qdrant/qdrant | Qdrant - High-performance, massive-scale Vector Database and Vector Search Engine for the next generation of AI. Also available in the cloud https://cloud.qdrant.io/ | 34,256 +93 this week | Apache-2.0 |
| 7 | chroma-core/chroma | Search infrastructure for AI | 29,182 +46 this week | Apache-2.0 |
| 8 | pgvector/pgvector | Open-source vector similarity search for Postgres | 22,818 +91 this week | not stated |
| 9 | weaviate/weaviate | Weaviate is an open-source vector database that stores both objects and vectors, allowing for the combination of vector search with structured filtering with the fault tolerance and scalability of a cloud-native database. | 16,759 +9 this week | BSD-3-Clause |
| 10 | Unstructured-IO/unstructured | Convert documents to structured data effortlessly. Unstructured is open-source ETL solution for transforming complex documents into clean, structured formats for language models. Visit our website to learn more about our enterprise grade Platform product for production grade workflows, partitioning, enrichments, chunking and embedding. | 15,362 +22 this week | Apache-2.0 |
Robotics
| Rank | Project | What it does | Stars | License |
|---|---|---|---|---|
| 1 | huggingface/lerobot | 🤗 LeRobot: Making AI for Robotics more accessible with end-to-end learning | 27,041 +164 this week | Apache-2.0 |
| 2 | google-deepmind/mujoco | Multi-Joint dynamics with Contact. A general purpose physics simulator. | 14,776 +128 this week | Apache-2.0 |
| 3 | rerun-io/rerun | Visualize, query, and stream to train on multimodal robotics data. | 11,372 +30 this week | Apache-2.0 |
| 4 | isaac-sim/IsaacLab | Unified framework for robot learning built on NVIDIA Isaac Sim | 7,995 +46 this week | BSD-3-Clause |
| 5 | NVIDIA/Isaac-GR00T | NVIDIA Isaac GR00T N1.7 - A Foundation Model for Generalist Robots. | 7,951 +48 this week | Apache-2.0 |
| 6 | ros2/ros2 | The Robot Operating System, is a meta operating system for robots. | 5,967 +30 this week | not stated |
| 7 | ros-navigation/navigation2 | ROS 2 Navigation Framework and System | 4,641 +21 this week | not stated |
| 8 | moveit/moveit2 | :robot: MoveIt for ROS 2 | 1,984 +12 this week | BSD-3-Clause |
Everyday Tools
| Rank | Project | What it does | Stars | License |
|---|---|---|---|---|
| 1 | AUTOMATIC1111/stable-diffusion-webui | Stable Diffusion web UI | 164,720 +84 this week | AGPL-3.0 |
| 2 | open-webui/open-webui | User-friendly AI Interface (Supports Ollama, OpenAI API, ...) | 150,335 +553 this week | not stated |
| 3 | Comfy-Org/ComfyUI | The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. | 130,597 +846 this week | GPL-3.0 |
| 4 | Mintplex-Labs/anything-llm | Stop renting your intelligence. Own it with AnythingLLM. Everything you need for a powerful local-first agent experience | 65,363 +217 this week | MIT |
| 5 | ggml-org/whisper.cpp | Port of OpenAI's Whisper model in C/C++ | 53,277 +134 this week | MIT |
| 6 | upscayl/upscayl | 🆙 Upscayl - #1 Free and Open Source AI Image Upscaler for Linux, MacOS and Windows. | 48,822 +86 this week | AGPL-3.0 |
| 7 | danny-avila/LibreChat | Enhanced ChatGPT Clone: Features Agents, MCP, Skills, DeepSeek, Anthropic, AWS, OpenAI, Responses API, Azure, Groq, o1, GPT-5, Mistral, OpenRouter, Vertex AI, Gemini, Artifacts, AI model switching, message search, Code Interpreter, langchain, DALL-E-3, OpenAPI Actions, Functions, Secure Multi-User Auth, Presets, open-source for self-hosting. Active | 42,594 +183 this week | MIT |
| 8 | jitsi/jitsi-meet | Jitsi Meet - Secure, Simple and Scalable Video Conferences that you use as a standalone app or embed in your web application. | 29,832 +34 this week | Apache-2.0 |
Posts on this topic
- AI in Practice
OpenAI tests Codex mode for permanently running AI agents
OpenAI released a code change in the public repository of the coding agent Codex on August 26, 2026, which prepares a new persistent mode. The mode allows the agent to continue working without a…
- Security
OpenAI: Agents execute code on 41 Hugging Face servers
OpenAI published a detailed technical report on the security incident at Hugging Face from July on August 26, 2026. The report reveals for the first time the full extent: around 700 AI agents…
- Security
Z.ai launches GLM-5.3-Flash – Cyber model remains locked
The Chinese AI company Z.ai released a new open language model, GLM-5.3-Flash, on August 26, 2026. The weights are available for free under the MIT license on Hugging Face, and the model processes up…
- AI-Economy
Nvidia negotiates purchase of Hugging Face for $12.9 billion
Nvidia is negotiating the purchase of the open-source platform Hugging Face for $12.9 billion, according to a report by the trade service The Information. The price is nearly three times the last…
- Security
DeepSeek: Chinese hackers double number of attacks
The Taiwanese security firm TeamT5 has determined that state-sponsored Chinese hacker groups have more than doubled their number of attacks since the routine deployment of AI chatbots. At least four…
- AI-Economy
Hugging Face is exploring sale for $13 billion
Hugging Face is reportedly exploring a sale that would value the open AI model platform at least $13 billion, according to a report by Business Insider. The company has hired a bank to gauge the…
- Research
Four new AI papers: Decoy Trick, Claude wins Gold, Agent RL
Today's selection from arXiv's new submissions of the past days connects four questions that rarely arise together: how to retroactively defend against remote safety bypasses of open models, how far…
- AI in Practice
Unsloth releases desktop app for local AI training
Unsloth AI released an open-source application called Unsloth Desktop on August 11, 2026, which consolidates local training and execution of AI models in a single app. The free software runs on…
- Security
Ray Flaw: CISA Reports Active Attacks on AI Computing Tool
The US cybersecurity agency CISA officially classified the vulnerability CVE-2025-62593 in the open-source AI framework Ray as actively exploited on August 17, 2026. A flaw in the browser defense…
- Security
Google Launches HEIR: A Compiler for Encrypted AI Computation
Google released the open-source compiler HEIR on August 14, 2026. The software transfers pre-trained AI models so that they can compute directly on encrypted data – servers process sensitive content…
- Security
GLM-5.3: Z.ai holds back model weights over cyber risk
The Chinese AI company Z.ai has introduced an open language model, GLM-5.3, which achieved 84.5 percent in the security benchmark CyberGym, narrowly ahead of Anthropic's Mythos 5 and OpenAI's GPT-5.6…
- Security
LiteLLM Leak: Hudson Rock Finds 153-Gigabyte Archive of Credentials
The security firm Hudson Rock has evaluated and released an archive of stolen credentials from the LiteLLM attack of March 2026. The 153-gigabyte collection contains 433,909 files, which analysts…
- Security
Taiwan confirms AI cyber attack: Justice Ministry also affected
Taiwan's Ministry of Digital Affairs has officially confirmed the AI-driven cyberattack on government systems uncovered in July and declared the investigation complete. New is the confirmation that…
- AI-Economy
River AI receives $1.1 billion for open AI models
The AI startup River AI has raised $1.1 billion in a combined seed and Series A round, led by investors General Catalyst and AMP PBC. Chip manufacturers Nvidia and AMD Ventures are also strategically…
- Security
Taiwan: Autonomous AI agents hack government systems
Chinese hackers have, according to a report by the Financial Times, for the first time carried out a largely autonomous cyberattack using AI agents against Taiwan's government. Up to eight agents…
- Security
LiteLLM attack hits over 2500 companies worldwide
A March 2026 attack on the open AI gateway LiteLLM has affected significantly more companies than previously known. The security firm CloudSEK now puts the number of affected companies at over 2500…
- AI in Practice
Meta releases Muse Glimmer: AI agent for a graphics card
Meta introduced an open agent model called Muse Glimmer with 30 billion parameters on August 10, 2026, which runs entirely locally on a single consumer graphics card. The company distills the model…
- Security
CrowdStrike: Criminals hijack stolen AI access
The security company CrowdStrike documents in its current Threat Hunting Report targeted attacks on corporate access to AI services. In a case recorded in May 2026, a group of attackers sent around…
- Research
Google DeepMind: WeatherNext warns a day earlier of hurricanes
Google DeepMind has introduced an AI model called WeatherNext that predicts cyclones one day earlier with the same accuracy as before. According to a study published in the journal Nature, three-day…
- Security
Kimi K3 bypasses cybersecurity test via GitHub
The Chinese AI model Kimi K3 broke out of a cybersecurity testing environment on August 7, 2026, as reported by the security company Frontier Security. Instead of solving a hacking task, the model…
- AI-Models
Xiaomi opens robotics AI with 100,000 training hours
The Chinese technology company Xiaomi has released its robotics foundational model Xiaomi-Robotics-1 as open-source software. Since August 3, code and model weights have been available on GitHub and…
- Security
Mistral Releases Shieldstral: AI Guardian for a 16-GB Graphics Card
Mistral released an open AI safety model called Shieldstral on August 4, 2026, which evaluates content against freely formulated rules – without retraining for new guidelines. The…
- AI-Policy
White House Excludes Open-Source AI from Security Tests
The White House discussed its months-prepared review regime for AI models on August 4, 2026, with leading tech companies – however, the framework will not be published. Only closed top models from…
- Security
GLM-5.2 narrows cyber gap to four to seven months
The British AI Security Institute (AISI) has compared open AI models in a new analysis with the cyber capabilities of leading systems such as Claude Opus 4.6. The analysis published on July 17 shows…
- AI in Practice
Linux Kernel Backs AI Tool Sashiko: Torvalds Rebuffs Critics
The Linux kernel now has a clear, official stance on artificial intelligence – and it comes straight from the top maintainer himself. In an email to the Linux kernel mailing list, Linus Torvalds…
- AI-Models
Kimi K3 reaches Anthropic level – local models catch up
Moonshot AI has today released Kimi K3 – the company's first large language model with open weights. With 2.8 trillion parameters and a million token context, K3 positions itself on par with…
- AI-Models
Thinking Machines releases open AI model Inkling
Thinking Machines Lab released its first publicly accessible language model on July 15, 2026. Inkling utilizes a Mixture-of-Experts architecture with 975 billion parameters, of which only 41 billion…
- AI-Economy
Nous Research aims for $1.5 billion valuation for Hermes
The AI startup Nous Research is reportedly negotiating a new funding round of at least $75 million, according to a report by TechCrunch. The valuation of the company behind the open AI agent Hermes…
- Research
Soofi S: German Consortium Releases Open AI Model
A German research consortium has released Soofi S, an open language model trained specifically for German and English. According to its developers, the 31.6-billion-parameter model surpasses the open…
- AI-Models
Robbyant releases AI world model LingBot-VA 2.0
The Ant Group subsidiary Robbyant has released LingBot-VA 2.0, an open AI model that translates camera images directly into robot movements. On the simulation benchmark RoboTwin 2.0, the system…
- AI-Economy
Ollama raises $65 million for open AI models
The provider Ollama announced a Series B funding round of $65 million on July 9, 2026. As Ollama states in its blog, the investment firm Theory Ventures is leading the round. For the company, founded…
- AI-Models
Tencent Hy3: 295B-MoE-Model with 21B Active Parameters
Tencent announced a fundamental rebuild of its pre-training and reinforcement learning infrastructure in 2026 – a rare signal for organizational change in an established AI lab. This decision was…
- AI
Mistral CEO Mensch warns against closed AI models
Mistral founder and CEO Arthur Mensch warns companies against relying on closed AI models from large providers. Providers of such models are increasingly storing customer data, thereby gaining deep…
- AI
pxpipe: How an Image Trick Lowers Claude Code Costs
An open-source tool called pxpipe exploits a peculiarity in Anthropic's pricing model: images are billed by pixel size, not by their text content. Developer Steven Chong has built a local proxy for…
Frequently asked questions
- What does open source mean for AI models?
- Open source means the blueprint is public. For AI models, the trained file – the so-called weights – is usually offered for download as well. That lets you run and adapt the model on your own machines, while closed models only run on the provider’s servers. One caveat: some providers call their model “open” but allow only limited use – it is worth reading the licence.
- Is open-source AI free?
- There is no licence fee for the software itself. The computing power still costs money: large models need expensive graphics cards or a cloud provider to run them. Smaller models, by contrast, run on an ordinary office computer. Free here means “no licence fee”, not “no effort”.
- Am I allowed to use open-source AI at work?
- Usually yes – what matters is the project’s licence. Licences such as Apache 2.0 or MIT explicitly permit commercial use, others restrict it. Anyone processing personal data should also clear the deployment with their own IT and data protection people. The advantage of open models: they can run inside the company, so the data never leaves the building.
- Where do the numbers on this page come from?
- From GitHub, the platform where most projects host their source code. A star there is a bookmark: developers mark the projects they find interesting. So a high star count shows attention, not quality. We fetch the numbers automatically; the date of each snapshot is shown above the tables.