CanMyPCRunAILocal AI compatibility
Check my PC

NVIDIA · 24GB VRAM

What AI models can a GeForce RTX 4090 run?

A GeForce RTX 4090 has 24GB of VRAM. Checked against 228 models from the Ollama library, 182 run fully on the GPU at 4K context and 22 more run with part of the model in system RAM.

182

run well

22

with trade-offs

228

models tested

Check my actual PC

Best AI models for a GeForce RTX 4090

Largest first — these fit entirely in 24GB of VRAM at 4K context, so the whole model is GPU-accelerated.

Models that run well on a GeForce RTX 4090
ModelBuildMemory neededVerdict
mistral-small3.1Mistral AI24B~21.0 GBGood fit
devstral-small-2Mistral AI24B~20.6 GBGood fit
mistral-small3.2Mistral AI24B~20.6 GBGood fit
lfm2Ollama Library24B~19.6 GBGood fit
devstralMistral AI24B~19.5 GBGood fit
magistralMistral AI24B~19.5 GBGood fit
gpt-ossOpenAI20B~18.8 GBGood fit
gpt-oss-safeguardOpenAI20B~18.8 GBGood fit
mistral-smallMistral AI22B~18.2 GBGood fit
codestralMistral AI22B~18.2 GBGood fit
solar-proUpstage22B~18.1 GBGood fit
phi4-reasoningMicrosoft14B~15.2 GBExcellent fit
deepseek-coder-v2DeepSeek AI16B~14.2 GBExcellent fit
deepseek-v2DeepSeek AI16B~14.2 GBExcellent fit
phi4Microsoft14B~12.4 GBExcellent fit
everythinglmOllama Library13B~10.9 GBExcellent fit
nexusravenOllama Library13B~10.9 GBExcellent fit
open-orca-platypus2Ollama Library13B~10.9 GBExcellent fit
codeupOllama Library13B~10.9 GBExcellent fit
wizard-vicunaOllama Library13B~10.9 GBExcellent fit
wizardlm-uncensoredOllama Library13B~10.9 GBExcellent fit
llama3.2-visionMeta AI11B~10.8 GBExcellent fit
mistral-nemoMistral AI12B~10.3 GBExcellent fit
gemma4Google DeepMinde2b-it-q4_K_M~9.9 GBExcellent fit
falcon2TII11B~9.5 GBExcellent fit

Showing the 25 largest of 182 models that run well.

How these results were calculated

Each model is evaluated at 4K context against 24GB of VRAM, counting model weights, KV cache, compute buffers and runtime overhead, minus a reserve for the display and operating system. Download sizes come from the official Ollama registry.

These figures assume 32GB of system RAM and working GPU drivers. Your own machine may differ — RAM, free disk space and whether an acceleration backend is actually installed all change the answer, which is what the PC scan measures directly.

GeForce RTX 4090 local AI — frequently asked questions

How many AI models can a GeForce RTX 4090 run?
Out of 228 models in the Ollama library, 182 run fully on a GeForce RTX 4090's 24GB of VRAM at 4K context, and a further 22 run with part of the model offloaded to system RAM, more slowly.
What is the largest AI model a GeForce RTX 4090 can run?
The heaviest build that fits entirely in VRAM is mistral-small3.1:24b (24B parameters), needing about 21.0 GB of memory from a 14.4 GB download.
Is 24GB of VRAM enough for local AI?
24GB is enough for 182 of the 228 models tested here, which covers most general-purpose and coding assistants. Very large models still need either partial CPU offload or a card with more memory.

Compare other graphics cards

GeForce RTX 5090GeForce RTX 5080GeForce RTX 5070 TiGeForce RTX 5070GeForce RTX 5060 Ti 16GBGeForce RTX 5060GeForce RTX 4080GeForce RTX 4070 TiGeForce RTX 4070GeForce RTX 4060 Ti 16GBGeForce RTX 4060 Ti 8GBGeForce RTX 4060GeForce RTX 3090GeForce RTX 3080GeForce RTX 3070GeForce RTX 3060 12GBGeForce RTX 3060 TiGeForce RTX 3050 8GB