Key takeaways
- Open source means the source is available under a licence that lets anyone use, study, modify and share it.
- Most 'open' AI models are open-weight: you can download and run the weights, but training data and code may not be released.
- Llama, Qwen, Mistral, Gemma and DeepSeek release open weights; GPT, Claude and Gemini are proprietary.
- n8n is source-available under a fair-code licence, not OSI open source; Next.js and Supabase are open source.
- Start contributing with documentation, tests and small bugs in a project you already use.
Open source software is software whose source code is published under a licence that lets anyone use, study, change and share it; Linux, Python, PostgreSQL, Git and VS Code's core are famous examples. In AI the term is contested: most models called open source are really open-weight, meaning you can download the trained weights, while the training data and full code stay private. This page lists essential open source software, explains open source AI, names the open models and agents, answers whether Google AI, n8n, Next.js and Supabase are open source, and shows how to contribute.
What is the best open source software?
The most valuable open source software for students and developers includes Linux, Git, Python, VS Code's open-source core, PostgreSQL, Docker's engine, Firefox, LibreOffice, VLC, GIMP and Blender, plus AI tools such as PyTorch, Hugging Face Transformers and Ollama.
Also asked as: what are the best open source software
By need:
- Development: Git, Python, Node.js, VS Code's core, PostgreSQL, Redis.
- AI and data: PyTorch, scikit-learn, Jupyter, Hugging Face libraries, Ollama, llama.cpp.
- Web: Next.js, React, Supabase, FastAPI.
- Everyday use: Firefox, LibreOffice, VLC, Thunderbird, Bitwarden.
- Creative: GIMP, Inkscape, Blender, OBS Studio, Audacity.
Much of the modern internet and nearly every AI product runs on open source. Learning to read and contribute to these projects is one of the best ways to grow as an engineer.
What is an open source AI model?
An open source AI model is one whose weights, and ideally code and training details, are published under a licence that lets anyone use, study, modify and share it. In practice, most models called open source are open-weight models with varying licence terms.
With an open-weight model you can download it, run it on your own hardware, fine-tune it with methods like QLoRA, and inspect its behaviour, which is not possible with closed models served only through APIs. Licences differ: some are fully permissive, such as Apache 2.0 or MIT, while others restrict commercial use above certain scales or certain uses. Always read the licence on the model card before building a product on a model.
What does open source mean in AI?
The Open Source Initiative's Open Source AI Definition says an AI system is open source when it grants the freedoms to use, study, modify and share it, which requires access to the model weights, the full code for training and running it, and enough information about the training data to reproduce a substantially equivalent system.
By that standard, few popular models qualify fully, because companies rarely release complete training data details. That is why careful writers say open-weight for models whose weights are downloadable but whose data and training pipelines are not. Fully open models from research groups do exist and publish data, code and checkpoints. For most engineering decisions, the practical questions are: can I download the weights, does the licence allow my use, and can I fine-tune and deploy it?
Which AI models are open source?
Widely used open-weight model families include Meta's Llama, Alibaba's Qwen, Mistral's open models, Google's Gemma, DeepSeek's models, Microsoft's Phi and OpenAI's open-weight releases. Licences range from permissive to custom terms with restrictions.
Popular choices by use:
- General chat and reasoning: Llama, Qwen, Mistral, DeepSeek models.
- Small and on-device: Gemma, Phi, small Llama and Qwen variants.
- Coding: coder variants of Qwen and others.
- Embeddings for RAG: open embedding models on Hugging Face.
- Speech: OpenAI's Whisper for transcription.
You can run them with Ollama, llama.cpp or vLLM, or through hosted providers. GPT, Claude and Gemini models are proprietary and only available through their companies' apps and APIs.
Is Google AI open source?
Partly. Google's Gemini models are proprietary, available through Google's apps and APIs. Google also releases Gemma, a family of open-weight models, and open-sources major tools such as TensorFlow and JAX.
Gemma models come in several sizes, including small ones that run on laptops and phones, and variants for specific uses. They are released under Google's Gemma terms, which allow broad use, including commercial use, with some conditions, so read the licence. Google researchers also published the original Transformer architecture, which underpins almost all modern language models. For students, Gemma is a good open model for experiments and fine-tuning practice, while Gemini is accessible through AI Studio's free tier.
Which AI agents are open source?
Open-source AI agents include coding agents such as OpenHands, Aider, Cline and the command-line agents Codex CLI and Gemini CLI, general agents such as Goose, and frameworks for building agents such as LangGraph, CrewAI and AutoGen.
Open-source agents let you inspect exactly what the agent does, run it with local models, and adapt it to your workflow. Many work with several model providers, so you can switch between hosted models and local ones through Ollama. Quality depends heavily on the model behind them: an open agent with a weak local model performs far worse than the same agent with a frontier model. Read the permissions model of any agent before letting it run commands on your machine.
Is n8n open source?
Not by the strict definition. n8n's source code is public and you can self-host it, but it uses the Sustainable Use License, a fair-code licence that restricts some commercial uses, such as offering n8n itself as a paid service. The Open Source Initiative does not recognise that as open source.
Also asked as: why is n8n open source
n8n chose this model to keep the code visible and free for internal use while protecting its business from competitors reselling it. For most users, including students, individuals and companies automating their own workflows, self-hosting n8n is allowed. If you plan to build a product that embeds or resells n8n, read the licence carefully or contact n8n about commercial terms. Fully open-source alternatives exist for some automation needs.
Is Next.js open source?
Yes. Next.js, the React framework maintained by Vercel, is open source under the MIT licence, and its code is on GitHub. You can use it for any project, commercial or not, and deploy it anywhere, not only on Vercel.
Vercel funds most of its development and offers hosting optimised for it, but self-hosting with Node.js or Docker is fully supported. The MIT licence only requires keeping the copyright notice. Next.js is a good project to learn from: its issues, discussions and examples show how a large open-source framework is run, and documentation fixes are a realistic first contribution.
Is Supabase open source?
Yes. Supabase, a backend platform built on PostgreSQL, is open source, with most of its code under the Apache 2.0 licence on GitHub. You can use the hosted service or self-host the stack with Docker.
Supabase combines open-source components: PostgreSQL for the database, PostgREST for an automatic REST API, a Realtime server, authentication, storage and edge functions, plus pgvector support for AI applications such as RAG. Self-hosting gives you control over data location and costs, but you take on maintenance, backups and security. For student projects, the hosted free tier is usually the quickest start; check its current limits.
How do I contribute to open source AI projects?
Pick a project you already use, read its contributing guide, start with documentation fixes, tests or issues labelled "good first issue", and open small, well-explained pull requests. Consistency matters more than size.
Also asked as: how to contribute to open source ai projects
A practical path:
- Use the project in a real task and note where you got stuck.
- Improve the docs or examples that confused you.
- Reproduce a reported bug and add a failing test.
- Fix small bugs, following the project's style and running its tests.
- Join the community channels and help answer questions.
Hugging Face libraries, LangChain, Ollama, vLLM and many model evaluation tools welcome contributions. Contributions on GitHub are public evidence of skill, and maintainers notice reliable contributors.
Sources
GenAI engineer and AI product builder with 2+ years shipping production-grade AI systems: RAG pipelines, fine-tuned LLMs, hybrid retrieval and multi-modal apps across vision, speech and OCR, architected end to end from ingestion to deployment. Leads engineering for SCULT INDIA's 14-member team, founded the 500+ member TechVerse Enclave community and has mentored 200+ students. Three hackathon first prizes: Changethon 2025 (IIT Roorkee), Product Genesis at Vividhotsava 2025 (CSJMU) and BYTEBATTLE (MeetKats).
Open to GenAI roles, hackathon judging, mentorship sessions and guest talks at colleges. For campus sessions the inviting institution covers the visit end to end: travel (a First AC train ticket for journeys up to five hours, a flight for anything longer), accommodation, meals, and pickup and drop; the honorarium and workshop fees are negotiable on WhatsApp or on the call, and requests are welcome from institutions that can cover the visit this way. Email pranjulrathour41@gmail.com.
