Weights
The individual learned numbers inside a model — “the weights” is shorthand for the whole trained brain you can download.
Think of it like
The recipe card after all the tweaking — hand someone the weights and they can cook the same dish.
Example
“Open weights” means you can download the numbers and run the model yourself.
How it actually works
Technically weights are one type of parameter (the multipliers on connections), but in practice people say “the weights” to mean the entire trained model file. Whether a lab releases them is the whole open-vs-closed debate: released weights can be run, inspected, and fine-tuned by anyone; withheld ones can only be reached through an API.
For product teams
Whether weights are public decides if you can self-host or are locked to a vendor’s API.
For engineers
The serialized tensor values of a trained network; “open weights” = distributable checkpoint.
Related
- Parameters — Another word for parameters.
- Fine-tuning — What fine-tuning nudges.
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