FitMyLLM

GPU memory calculator·byKarthic

EasyAdvanced

Find out how much GPU memory your AI model aka LLM (Large Language Model) needs before you buy hardware or rent cloud GPUs.

LLM Model Parameter

1. Choose a model family

?

2. Choose a size

?

Precision?

Batch size?

Safety margin (%)?

Context length?

4K tokens

Training mode?

Running the model only (no learning).

Off

Memory Estimate

?

23.3 GBinference

✅ Fits on NVIDIA RTX 3090 24 GB0.7 GB spare

Weights

Weights18.4 GB77%

KV cache2.5 GB10%

Activations1.2 GB5%

Overhead1.1 GB5%

Spare0.7 GB3%

Hardware Recommendations?

NVIDIA A100 40 GB✅

1x40.0 GB

58%

NVIDIA A6000 48 GB✅

1x48.0 GB

49%

NVIDIA L40S 48 GB✅

1x48.0 GB

49%

NVIDIA A100 80 GB✅

1x80.0 GB

29%

NVIDIA H100 80 GB✅

1x80.0 GB

29%

NVIDIA H200 141 GB✅

1x141.0 GB

17%

NVIDIA B200 192 GB✅

1x192.0 GB

12%

NVIDIA B300 288 GB✅

1x288.0 GB

8%