GPUs & Macs › NVIDIA GTX 1660
NVIDIA GTX 1660: local LLM speed
Llama 2 7B (Q4_0) generates 41.4 tokens/s on the NVIDIA GTX 1660 (CUDA), and processes prompts at 154 tokens/s.
Measured results
| Model | Generate tok/s | Prompt tok/s | Quant | Backend | Test | Source |
|---|---|---|---|---|---|---|
| Llama 2 7B | 41.4 | 154 | Q4_0 | CUDA FA on | tg128 | llama.cpp #15013 details |
| Llama 2 7B | 41.4 | 149 | Q4_0 | CUDA FA off | tg128 | llama.cpp #15013 details |
| Llama 3.2 1B Instruct | 34.2 | 765 | Q4_K_M | llamafile | LocalScore avg | LocalScore details |
| Llama 3.1 8B Instruct | 9.0 | 129 | Q4_K_M | llamafile | LocalScore avg | LocalScore details |
What fits on the NVIDIA GTX 1660
Estimates for this chip are scaled from a measured run: Llama 2 7B Q4_0 at 41.4 tok/s (CUDA, source). Fit assumes the whole model in GPU memory with ~8K context.
| Model | Best quant that fits | File | Expected tok/s |
|---|---|---|---|
| gemma-4-12B-it | UD-IQ2_M | 4.2 GB | 37.6 estimate |
| Qwen3.5-9B | UD-IQ3_XXS | 4.0 GB | 39.4 estimate |
| Qwen3.5-4B | Q8_0 | 4.5 GB | 35.3 estimate |
| MiMo-V2.6-Distill-Qwen-9B | Q3_K_M | 4.5 GB | 35.4 estimate |
| gemma-4-E4B-it | Q3_K_M | 4.1 GB | 39.0 estimate |
| Qwen3-4B | Q8_0 | 4.3 GB | 37.0 estimate |
| gemma-3-1b-it | Q8_0 | 1.1 GB | 132 estimate |
| gemma-4-E2B-it | Q4_0 | 2.8 GB | 55.7 estimate |
| DeepSeek-R1-0528-Qwen3-8B | Q3_K_M | 4.1 GB | 38.4 estimate |
| Llama-3.2-3B-Instruct | Q8_0 | 3.4 GB | 46.3 estimate |
| Phi-4-mini-instruct | Q8_0 | 4.1 GB | 38.8 estimate |
| gemma-3-4b-it | Q8_0 | 4.1 GB | 38.3 estimate |
| Qwen3-14B | UD-IQ1_M | 4.1 GB | 39.0 estimate |
| gemma-3-12b-it | UD-IQ2_M | 4.4 GB | 36.3 estimate |
Similar hardware
NVIDIA GTX 1060 · NVIDIA RTX 2060 · NVIDIA Quadro RTX 3000 · NVIDIA GTX 1660 SUPER · NVIDIA P106-090 · NVIDIA GTX 1660 Ti · NVIDIA RTX 2070 SUPER · NVIDIA RTX 4060 Ti