GPUs & Macs › NVIDIA RTX 4070 Ti SUPER
NVIDIA RTX 4070 Ti SUPER: local LLM speed
Llama 2 7B (Q4_0) generates 136 tokens/s on the NVIDIA RTX 4070 Ti SUPER (Vulkan), and processes prompts at 6,801 tokens/s.
Measured results
| Model | Generate tok/s | Prompt tok/s | Quant | Backend | Test | Source |
|---|---|---|---|---|---|---|
| Llama 2 7B | 136 | 6,801 | Q4_0 | Vulkan FA on | tg128 | llama.cpp #10879 details |
| Llama 2 7B | 133 | 7,612 | Q4_0 | CUDA FA on | tg128 | llama.cpp #15013 details |
| Llama 2 7B | 132 | 6,925 | Q4_0 | CUDA FA off | tg128 | llama.cpp #15013 details |
| Llama 2 7B | 129 | 6,099 | Q4_0 | Vulkan FA off | tg128 | llama.cpp #10879 details |
| Llama 3.2 1B Instruct | 161 | 13,268 | Q4_K_M | llamafile | LocalScore avg | LocalScore details |
| Llama 3.1 8B Instruct | 53.9 | 3,561 | Q4_K_M | llamafile | LocalScore avg | LocalScore details |
| Qwen2.5 14B Instruct | 36.8 | 2,147 | Q4_K_M | llamafile | LocalScore avg | LocalScore details |
What fits on the NVIDIA RTX 4070 Ti SUPER
Estimates for this chip are scaled from a measured run: Llama 2 7B Q4_0 at 136 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-IQ3_XXS | 10.9 GB | 47.5 estimate |
| gemma-4-12B-it | Q8_0 | 12.7 GB | 41.0 estimate |
| Qwen3.5-9B | Q8_0 | 9.5 GB | 54.5 estimate |
| Qwen3.5-4B | Q8_0 | 4.5 GB | 116 estimate |
| Qwen3.6-35B-A3B | UD-IQ3_XXS | 13.2 GB | 229 estimate |
| gemma-4-26B-A4B-it | UD-IQ4_XS | 13.6 GB | 124 estimate |
| MiMo-V2.6-Distill-Qwen-9B | Q8_0 | 9.5 GB | 54.4 estimate |
| gemma-4-E4B-it | Q8_0 | 8.2 GB | 63.4 estimate |
| Muse-Glimmer-30B | UD-IQ3_XXS | 13.1 GB | 39.5 estimate |
| Qwen3-4B | Q8_0 | 4.3 GB | 121 estimate |
| gemma-4-31B-it | UD-IQ3_XXS | 11.8 GB | 43.9 estimate |
| gemma-4-12B-it-qat-q4_0-unquantized | UD-Q4_K_XL | 6.7 GB | 77.3 estimate |
| gemma-3-1b-it | Q8_0 | 1.1 GB | 433 estimate |
| gemma-4-E2B-it | Q8_0 | 5.0 GB | 104 estimate |
Similar hardware
NVIDIA RTX 5080 · NVIDIA RTX 5070 Ti · NVIDIA RTX 4080 SUPER · NVIDIA RTX 4080 · NVIDIA RTX 5060 Ti · NVIDIA RTX PRO 2000 Blackwell · NVIDIA Quadro P5000 · NVIDIA Quadro RTX 5000