United States
What AI models can a PNY VCNRTX4000ADA-PB run?
With 20 GB of VRAM and 32 GB of system RAM, here's how this card handles local AI models (Q4 quantization, up to 8K context).
Upgrade tip: Your GPU's VRAM is the main limit. A 24 GB card would run 4 more models fully on the GPU.
Runs well
18-
Llama 3.1 8B
Needs ~7 GB
-
Qwen2.5 7B
Needs ~6 GB
-
Qwen2.5-Coder 7B
Needs ~6 GB
-
FLUX.1 dev (12B)
Needs ~13 GB
-
Qwen2.5 14B
Needs ~12 GB
-
DeepSeek-R1 Distill Qwen 14B
Needs ~12 GB
-
DeepSeek-R1 Distill Qwen 7B
Needs ~6 GB
-
Mistral Small 24B
Needs ~17 GB
-
FLUX.1 schnell (12B)
Needs ~13 GB
-
Mistral 7B
Needs ~7 GB
-
Phi-4 (14B)
Needs ~12 GB
-
Stable Diffusion XL 1.0
Needs ~5 GB
-
Gemma 2 9B
Needs ~10 GB
-
Llama 3.2 3B
Needs ~4 GB
-
Stable Diffusion 3.5 Large
Needs ~10 GB
-
Qwen2.5 3B
Needs ~4 GB
-
Phi-3 Mini (3.8B)
Needs ~5 GB
-
Gemma 2 2B
Needs ~4 GB
Runs with compromises
8-
Qwen2.5 32B
Needs ~23 GB
-
Qwen2.5-Coder 32B
Needs ~23 GB
-
DeepSeek-R1 Distill Qwen 32B
Needs ~23 GB
-
Llama 3.3 70B
Needs ~45 GB
-
DeepSeek-R1 Distill Llama 70B
Needs ~45 GB
-
Qwen2.5 72B
Needs ~46 GB
-
Mixtral 8x7B
Needs ~30 GB
-
Gemma 2 27B
Needs ~20 GB
Needs more hardware
0- None.