RTX 5090
83 AI models fit in 32 GB VRAM at Q4 native. 8 more run with CPU offloading. Real benchmarks below.
llama.cpp 0.2.x · CUDA 12 · Ubuntu 22.04 · Prices verified on Amazon · methodology →
Execution Context
Check current offer
Amazon affiliate link for RTX 5090
Some links are Amazon affiliate links. We may earn a commission at no extra cost to you. Amazon cookies may last up to 24 hours after your click.
Full Specifications
NVIDIA · 2025-01
| VRAM | 32 GB GDDR7 |
| Bandwidth | 1792 GB/s |
| FP16 TFLOPS | 105 |
| AI Score | 140 / 140 |
| CUDA Cores | 21,760 |
| Tensor Cores | 680 |
| TDP | 575 W |
| PCIe | Gen 5 |
| Slots | 3 |
| Power Connector | 16-pin |
| Price Band | Flagship |
| Released | 2025-01 |
AI Benchmarks
Real inference measurements — llama.cpp Q4_K_M
| Task | Result |
|---|---|
| Llama 1B Q4 | 400 tok/s |
| Llama 3B Q4 | 200 tok/s |
| Llama 7B Q4 | 155 tok/s |
| Llama 13B Q4 | 84 tok/s |
| Llama 30B Q4 | 42 tok/s |
| Llama 70B Q4 | Offload or multi-GPU |
| Stable Diffusion 512px | 1.2s / img |
| Whisper Large RTF | 0.2x |
RTF < 1.0 = faster than real time. For Stable Diffusion and Whisper lower is better; for tokens/s higher is better.
Compare RTX 5090 with another GPU
Is an upgrade worth it? Compare specs and real benchmarks side by side.
Open comparator →Compatible AI Models — RTX 5090
83 models run fully in VRAM · 8 with CPU offloading
Flux.1 Dev
Whisper Large V3
Stable Diffusion 3.5 Large
FLUX.2 Dev
Qwen2.5-Coder 32B
DeepSeek R1 Distill 32B
Qwen2.5 32B
Stable Diffusion 3.5 Medium
Show all 83 compatible models →
Also runs with CPU offloading (8)
- Llama 3.3 70B 21 GB Q2 How to install →
- Qwen2.5 72B 21 GB Q2 How to install →
- Llama 3.1 70B 20 GB Q2 How to install →
- Command R+ 30 GB Q2 How to install →
- Qwen3-Coder-Next 80B-A3B 22 GB Q2 How to install →
- Llama 3.2 90B Vision 27 GB Q2 How to install →
- Llama 4 Scout 30 GB Q2 How to install →
- Hermes 3 70B 20 GB Q2 How to install →
RTX 5090 · Amazon
GPU pricing moves frequently across retailers. Check the current offer before you buy.
Check current offerSome links are Amazon affiliate links. We may earn a commission at no extra cost to you. Amazon cookies may last up to 24 hours after your click.
RTX 5090 for Local AI
La RTX 5090 con 32GB de GDDR7 es una GPU de alto rendimiento para IA local. Puede ejecutar modelos de 7B, 13B y hasta 70B parámetros en cuantización Q2, cubriendo la mayoría de los casos de uso prácticos: asistentes de chat, generación de código, análisis de texto y generación de imágenes con Stable Diffusion.
En benchmarks reales, la RTX 5090 alcanza 155 tokens/segundo en Llama 7B Q4 — velocidad suficiente para conversación fluida en tiempo real. Los modelos de 13B como Phi-4 o Qwen2.5 14B corren cómodamente en Q4, y modelos 70B son accesibles via Q2 con 8 opciones.
La relación rendimiento-precio de la RTX 5090 la hace una de las mejores opciones para usuarios técnicos que quieren IA local seria. Usa la calculadora de VRAM para ver exactamente qué cuantización necesita cada modelo, o consulta nuestra guía para elegir GPU.
Plan your full AI build
RTX 5090 · 32 GB VRAM — configure PSU, RAM, storage and check compatible models.
Related articles
Not sure which model to run on your RTX 5090?
The VRAM calculator tells you exactly which quantization you need.
Get the best price for RTX 5090
Open Amazon with our affiliate link and check availability, variants, and current deals.