⚡ Model Size Converter

Parameters → GB · training memory estimator

e.g., 7B, 0.5B, 405B
Optimizer
Display unit:
14.00 GB
Model weights
84.00 GB
Optimizer memory
14.00 GB
Gradient memory
112.00 GB
Total training
112.00 GB
Min GPU VRAM
H100 80GB ×2
Recommended GPU

📊 Quick reference — common model sizes (FP16, Adam, gradients)

ParametersWeightsOptimizerGradientsTotal trainMin VRAM

* Assumes FP16 (2B), Adam optimizer, gradients included. Values in GB.