Open Source
Software or AI models whose underlying code or weights are publicly available for anyone to inspect, modify, and use.
Open source means the source code (for software) or model weights (for AI) are publicly released under a licence that permits inspection, modification, and redistribution. The opposite is closed source or proprietary — where code or weights are private (as with GPT-4o or Claude).
Open source in AI refers to two things: 1. Open weights: The trained model parameters are released — anyone can download and run the model locally. Examples: Meta's Llama, Mistral, Stable Diffusion, Cactus Needle. 2. Open code: The training scripts, fine-tuning code, or application code are available on GitHub.
Why it matters for builders: - No API costs: Run the model locally for zero marginal cost per call. - No rate limits: Not subject to provider usage policies or uptime. - Customisable: Fine-tune on your own data without asking permission. - Privacy: Data never leaves your infrastructure.
Licences vary: MIT (most permissive — commercial use freely allowed), Apache 2.0 (similar), and community licences (some restrict commercial use — check before shipping).
Example
Stable Diffusion is open source — you download the weights, run it on your own GPU, generate unlimited images, and build commercial products on top of it. DALL-E 3 is closed source — you pay OpenAI per image via API.
Related terms
Large Language Model (LLM)
A neural network trained on massive amounts of text to predict and generate human language.
Fine-Tuning
The process of updating a pre-trained model with task-specific or domain-specific data to improve performance.
Inference
The process of a trained model generating a response to an input. When you chat with ChatGPT, that's inference.
GPU
Graphics Processing Unit — a chip originally built for rendering graphics, now the primary hardware for training and running AI models due to its ability to perform thousands of calculations simultaneously.