A→Z
A2ZAI
Back to Glossary
concepts

Open Weights

AI models where the trained parameters are publicly released, enabling local deployment and modification.

Share:

Definition

Open weights models have their trained model parameters publicly available for download, allowing anyone to run them locally or fine-tune them.

Open Weights vs Open Source: - Open weights: Parameters available - Open source: Parameters + training code + data - Many "open" models are weights-only

Benefits: - Run locally (privacy) - No API costs - Customization via fine-tuning - Research and experimentation - No vendor lock-in

Notable Open Weights Models: - Meta Llama 3 family - Mistral models - Qwen models - Gemma (Google) - Phi models (Microsoft)

Considerations: - Require compute to run - May have license restrictions - Large file sizes - Self-hosting complexity

Examples

Downloading Llama 3.1 70B to run on your own GPU cluster without sending data to external APIs.

Want more AI knowledge?

Get bite-sized AI concepts delivered to your inbox.

Free daily digest. No spam, unsubscribe anytime.

Discussion