Rig score · nvidia
Nvidia GeForce RTX 5090 (32GB)
Runs 108 of 135 models (80%)
Excellent, runs up to ~32B comfortably.
Biggest it can run: Llama-3.3-Nemotron-Super-49B-v1
- Grade
- A
- Runnable
- 108
- Total
- 135
- Coverage
- 80%
What Nvidia GeForce RTX 5090 (32GB) can run
At 32 GB of VRAM, Nvidia GeForce RTX 5090 (32GB) runs 108 of 135 tracked local models, 80% coverage that earns a A. The largest it handles is Llama-3.3-Nemotron-Super-49B-v1 at 49B parameters; anything bigger needs more memory or a lower quantization.
Only the 32 GB of VRAM drives speed: a model that spills into system RAM slows sharply. Keep the weights plus KV cache under VRAM and prefer Q4_K_M to leave headroom.
Top models on Nvidia GeForce RTX 5090 (32GB)
FAQ
What local AI rig score does Nvidia GeForce RTX 5090 (32GB) get?
Nvidia GeForce RTX 5090 (32GB) scores A: it runs 108 of 135 tracked local LLMs (80%). The biggest it can run is Llama-3.3-Nemotron-Super-49B-v1.
How big a model can Nvidia GeForce RTX 5090 (32GB) run?
Nvidia GeForce RTX 5090 (32GB) runs models up to Llama-3.3-Nemotron-Super-49B-v1 (49B parameters) at Q4_K_M. Speed depends on keeping the whole model inside VRAM.
Is Nvidia GeForce RTX 5090 (32GB) good for running local LLMs?
Only the 32 GB of VRAM drives speed: a model that spills into system RAM slows sharply. Keep the weights plus KV cache under VRAM and prefer Q4_K_M to leave headroom.
Want your own? Check your rig on the homepage →
Score from coverage of 135 tracked models, estimated with our memory methodology.