Open-Weight vs. Open-Source vs. Closed AI Models
New'Open-source LLM' is widely used for models that are only open-weight. Open-weight models (Llama, Mistral, Qwen, DeepSeek, Gemma) let anyone download and run the trained model, sometimes with licence restrictions, but don't publish the training data or full recipe. Truly open-source models release those too, and are rare. Closed models (GPT, Claude, Gemini) are only available through the provider's API. Running open-weight models in-house needs GPU and serving skills that API users never touch.
How to tell them apart on a resume
Open-weight
Llama, Mistral, Qwen, DeepSeek, Gemma, Hugging Face, vLLM, Ollama, self-hosting, quantization — the team runs the model on its own hardware or cloud.
Fully open-source
OLMo, Pythia, published training data and code, research or academic settings — rare in industry roles.
Closed (API-only)
OpenAI, Anthropic, Gemini, Azure OpenAI, AWS Bedrock — the model runs on the provider's servers and is called over an API.
The question that settles it
“Did you run the model on your own infrastructure, or call it through a provider's API — and what made you choose that?”
Read the full definitions
Open the full tool for the other look-alike pairs, role profiles, and the JD decoder.