BTC
ETH
SOL
XRP
BNB
TON
DOT
LTC
POL
HBAR
NEAR
ARB
UNI
The numbers inside an AI model that it learned in training. Model size is counted in parameters, often billions of them.
AI Drafted with AI help · Reviewed by Keshav Aggarwal · Updated
In simple words
Imagine a huge music mixing desk with billions of little knobs. Training turns each knob a tiny bit until the music sounds right.
Those knobs are the parameters.
The technical side
Parameters are a neural network’s weights and biases, set during training. Meta’s Llama 3.1 405B, released in July 2024, has 405 billion; many useful models have 7 to 70 billion. More parameters usually mean more capability but also more memory, slower answers and higher cost.
Mixture-of-experts models only use part of their parameters for each token, so their total size overstates the work done per answer.
Parameters kya hai?
Parameters model ke andar ke numbers hain jo training mein seekhe jaate hain. Model kitna bada hai, ye parameters se naapa jaata hai.
Related terms
- LLM (Large Language Model)AIAn AI model trained on huge amounts of text so it can read and write language, like the models behind ChatGPT, Claude and Gemini.
- Fine-TuningAITraining an existing AI model a little more on your own examples, so it gets better at one task or speaks in one style.
- QuantisationAIShrinking an AI model by storing its numbers with less precision, so it runs faster and on smaller hardware with a small loss of quality.