Open Source AI

Definition

AI models released with open licenses (MIT, Apache 2.0) allowing anyone to use, modify and deploy them.

Why It Matters

Open-source models can be self-hosted, audited, fine-tuned and run forever without depending on a vendor's API staying up or its pricing staying reasonable. That's why every serious AI platform (including this one) keeps an open-source path.

Key Points

• Apache 2.0 and MIT licenses allow commercial use, modification and redistribution with attribution, the two most permissive standard licenses.
• Meta's Llama 3 license restricts commercial use for companies with more than 700M monthly active users, not fully open by the Open Source Initiative definition.
• GGUF format (used by llama.cpp) enables CPU inference on consumer hardware, a quantised 7B model runs at ~30 tokens/second on an Apple M2.
• Hugging Face Hub is the primary distribution point: over 1 million model repositories and 200K+ dataset repositories as of 2025.
• Open-source model quality: Qwen 2.5 72B and Llama 3.1 70B now match GPT-4-class performance on many benchmarks while being self-hostable.

Example

Llama, Qwen, Mistral, DeepSeek, FLUX, Stable Diffusion, Whisper, all Apache 2.0 or MIT licensed. You can download the weights, run them on your own hardware, modify them, and ship products on top without paying anyone.

Related Terms

• LLM (Large Language Model)A neural network trained on massive text datasets that can generate, understand and manipulate human language. Examples: GPT-4, Qwen, Claude.
• Fine-TuningTraining a pre-trained AI model on specialized data to improve performance on specific tasks.
• ParameterA trainable weight in an AI model. Larger models have more parameters (7B, 70B, 400B).

Quick Facts