GPUs & Macs › NVIDIA RTX 3090
NVIDIA RTX 3090: local LLM speed
Llama 2 7B (Q4_0) generates 172 tokens/s on the NVIDIA RTX 3090 (Vulkan), and processes prompts at 5,201 tokens/s.
Measured results
| Model | Generate tok/s | Prompt tok/s | Quant | Backend | Test | Source |
|---|---|---|---|---|---|---|
| Llama 2 7B | 172 | 5,201 | Q4_0 | Vulkan FA on | tg128 | llama.cpp #10879 details |
| Llama 2 7B | 164 | 4,666 | Q4_0 | Vulkan FA off | tg128 | llama.cpp #10879 details |
| Llama 2 7B | 162 | 5,560 | Q4_0 | CUDA FA on | tg128 | llama.cpp #15013 details |
| Llama 2 7B | 158 | 5,175 | Q4_0 | CUDA FA off | tg128 | llama.cpp #15013 details |
| Llama 3.2 1B Instruct | 270 | 13,482 | Q4_K_M | llamafile | LocalScore avg | LocalScore details |
| gpt-oss 20B | 162 | 5,171 | MXFP4 | CUDA FA on | tg128 | llama.cpp #15396 details |
| Llama 3.1 8B Instruct | 96.3 | 3,550 | Q4_K_M | llamafile | LocalScore avg | LocalScore details |
| Qwen2.5 14B Instruct | 56.2 | 1,950 | Q4_K_M | llamafile | LocalScore avg | LocalScore details |
What fits on the NVIDIA RTX 3090
Estimates for this chip are scaled from a measured run: Llama 2 7B Q4_0 at 172 tok/s (Vulkan, source). Fit assumes the whole model in GPU memory with ~8K context.
| Model | Best quant that fits | File | Expected tok/s |
|---|---|---|---|
| Qwen3.8-27B | UD-Q6_K | 22.0 GB | 29.8 estimate |
| gemma-4-12B-it | Q8_0 | 12.7 GB | 51.8 estimate |
| Qwen3.5-9B | Q8_0 | 9.5 GB | 68.9 estimate |
| Qwen3.5-4B | Q8_0 | 4.5 GB | 146 estimate |
| Qwen3.6-35B-A3B | UD-Q4_K_S | 20.9 GB | 183 estimate |
| gemma-4-26B-A4B-it | UD-Q5_K_M | 21.2 GB | 101 estimate |
| MiMo-V2.6-Distill-Qwen-9B | Q8_0 | 9.5 GB | 68.7 estimate |
| gemma-4-E4B-it | Q8_0 | 8.2 GB | 80.1 estimate |
| Muse-Glimmer-30B | UD-Q5_K_M | 19.2 GB | 34.2 estimate |
| Qwen3-4B | Q8_0 | 4.3 GB | 153 estimate |
| gemma-4-31B-it | Q5_K_M | 21.7 GB | 30.3 estimate |
| gemma-4-26B-A4B-it-qat-q4_0-unquantized | UD-Q4_K_XL | 14.2 GB | 150 estimate |
| gemma-4-12B-it-qat-q4_0-unquantized | UD-Q4_K_XL | 6.7 GB | 97.7 estimate |
| gemma-3-1b-it | Q8_0 | 1.1 GB | 547 estimate |
Similar hardware
NVIDIA RTX 4090 D · NVIDIA RTX 4090 · NVIDIA RTX 3090 Ti · NVIDIA TITAN RTX · NVIDIA Quadro P6000 · NVIDIA RTX 5090 · NVIDIA RTX 5080 · NVIDIA RTX 5070 Ti