open source AI
Models released with weights publicly available, usually under permissive licenses, enabling anyone to run, modify, or redistribute them.
Think of it like
The difference between owning a car and renting one—open-source models are yours to tinker with.
Example
Llama, Mistral, Gemma, OLMo. Download them, fine-tune them on your data, deploy on your hardware, modify the architecture.
How it actually works
Open-source AI is a spectrum—some models are research releases, others are production-ready. They come with training data, architecture details, and sometimes training code. Tradeoff: full control and transparency, but you own the compute to run and update them. Closed models are often better, but OS models are catching up.
For product teams
Reduces lock-in and enables custom fine-tuning; shifts cost from API to infrastructure.
For engineers
Full stack control; can optimize for latency, privacy, cost.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome