Best local AI models for GeForce RTX 5090
32 GB of GDDR7 changes what fits: a 32B model at Q5_K_M with real context, or a 14B QLoRA run with room to spare.
Specification
VRAM32 GB
Usable for models31 GB
System RAM64 GB
Bandwidth1792 GB/s
Local trainingSupported
Operating systemswindows, linux
Runs well
29 models fit this machine
Estimated
| Model | Params | Quantization | Memory | ~tok/s | Fit | Fine-tune |
|---|---|---|---|---|---|---|
| Qwen2.5 Coder 32B Instruct | 32.8B | Q4_K_M | 21 GB | 70 | Good fit | Yes |
| Qwen2.5 32B Instruct | 32.8B | Q4_K_M | 21 GB | 70 | Good fit | Yes |
| Phi-4 14B | 14.7B | Q8_0 | 17 GB | 89 | Excellent fit | Yes |
| DeepSeek-R1-Distill-Qwen-32B | 32.8B | Q4_K_M | 21 GB | 70 | Good fit | Yes |
| Mistral Small 24B Instruct | 23.6B | Q5_K_M | 18 GB | 83 | Excellent fit | Yes |
| Qwen2.5 14B Instruct | 14.8B | Q8_0 | 17 GB | 88 | Excellent fit | Yes |
| Gemma 3 27B Instruct | 27.4B | Q5_K_M | 24 GB | 71 | Good fit | Yes |
| Qwen3 30B-A3B | 30.5B | Q5_K_M | 22 GB | 592 | Good fit | Yes |
| StarCoder2 15B | 16B | Q8_0 | 17 GB | 81 | Excellent fit | Yes |
| Qwen3 14B | 14.8B | Q8_0 | 17 GB | 88 | Excellent fit | Yes |
| DeepSeek-R1-Distill-Qwen-14B | 14.8B | Q8_0 | 17 GB | 88 | Excellent fit | Yes |
| Mistral Nemo 12B Instruct | 12.2B | Q8_0 | 14 GB | 107 | Excellent fit | Yes |
| Qwen2.5 Coder 7B Instruct | 7.6B | F16 | 15 GB | 91 | Excellent fit | Yes |
| Gemma 3 12B Instruct | 12.2B | Q8_0 | 16 GB | 107 | Excellent fit | Yes |
| DeepSeek-Coder-V2-Lite Instruct | 15.7B | Q8_0 | 18 GB | 543 | Excellent fit | Yes |
| Qwen2.5 7B Instruct | 7.6B | F16 | 15 GB | 91 | Excellent fit | Yes |
| Llama 3.1 8B Instruct | 8B | F16 | 16 GB | 87 | Excellent fit | Yes |
| Gemma 2 9B Instruct | 9.2B | F16 | 20 GB | 75 | Good fit | Yes |
Fine-tunable here
Models you can train on this machine
Qwen2.5 Coder 32B Instruct
QLORA · 22 GB peak
Qwen2.5 32B Instruct
QLORA · 22 GB peak
Phi-4 14B
QLORA · 12 GB peak
DeepSeek-R1-Distill-Qwen-32B
QLORA · 22 GB peak
Mistral Small 24B Instruct
QLORA · 17 GB peak
Qwen2.5 14B Instruct
QLORA · 12 GB peak
Gemma 3 27B Instruct
QLORA · 19 GB peak
Qwen3 30B-A3B
QLORA · 18 GB peak