Parameter
A parameter is one of the internal numbers a trained model stores: billions of them, each tuned during training, and together they encode everything the model learned.
No jargon, no math
In plain words, what is it?
Think of parameters as the model's dials, one per connection the training settled on. Your input arrives as numbers, flows through the network's layers, and at each layer it is multiplied by matrices full of parameters, summed, and reshaped. Every parameter nudges the answer up or down a little; billions of tiny nudges compose a fluent reply. The count is the model's headline size: a “7B” model holds 7 billion parameters. That number drives everything operational: more parameters mean more VRAM to hold them and more compute per token. Quantization shrinks each parameter from 16 to 8 or 4 bits so the same model fits a smaller GPU, at a small cost in precision.
Terms that live nearby.
Training
Training is the process that turns an untrained model into a working AI: you feed it thousands or millions of examples, compare its guesses with the correct answers, and nudge its internal settings until the guesses get good.
ReadQuantization
Quantization is a compression technique that rewrites a model's weights from high-precision decimals (32-bit floats) to smaller numbers (8-bit or 4-bit integers), so the model takes up less memory and runs faster on cheaper hardware, at a small, measurable cost in accuracy.
ReadGPU
A GPU is a processor built to run thousands of small calculations at the same time. Designed for graphics, it turned out to be the perfect engine for AI, because running a neural network is mostly one enormous pile of matrix multiplications.
Read§ 03
Want this running in your stack?
We design and ship AI systems with permissions, logs and human approval. It starts with a 30-minute scoping call.