GPUs & Macs › NVIDIA L40S
NVIDIA L40S: local LLM speed
3 measured results for the NVIDIA L40S.
Measured results
| Model | Generate tok/s | Prompt tok/s | Quant | Backend | Test | Source |
|---|---|---|---|---|---|---|
| Llama 3.2 1B Instruct | 238 | 21,532 | Q4_K_M | llamafile | LocalScore avg · n=2 | LocalScore LocalScore details |
| Llama 3.1 8B Instruct | 45.9 | 5,328 | Q4_K_M | llamafile | LocalScore avg | LocalScore details |
| Qwen2.5 14B Instruct | 19.3 | 2,868 | Q4_K_M | llamafile | LocalScore avg | LocalScore details |
What fits on the NVIDIA L40S
Estimates for this chip are scaled from a measured run: Llama 3.1 8B Instruct Q4_K_M at 45.9 tok/s (llamafile, source). Fit assumes the whole model in GPU memory with ~8K context.
| Model | Best quant that fits | File | Expected tok/s |
|---|---|---|---|
| Qwen3.8-27B | Q8_0 | 29.0 GB | 7.8 estimate |
| gemma-4-12B-it | Q8_0 | 12.7 GB | 17.8 estimate |
| Qwen3.5-9B | Q8_0 | 9.5 GB | 23.7 estimate |
| Qwen3.5-4B | Q8_0 | 4.5 GB | 50.4 estimate |
| Qwen3.6-35B-A3B | Q8_0 | 36.9 GB | 35.7 estimate |
| gemma-4-26B-A4B-it | Q8_0 | 26.9 GB | 27.3 estimate |
| MiMo-V2.6-Distill-Qwen-9B | Q8_0 | 9.5 GB | 23.7 estimate |
| gemma-4-E4B-it | Q8_0 | 8.2 GB | 27.6 estimate |
| Muse-Glimmer-30B | Q8_0 | 29.6 GB | 7.6 estimate |
| Qwen3-4B | Q8_0 | 4.3 GB | 52.8 estimate |
| gemma-4-31B-it | Q8_0 | 32.6 GB | 6.9 estimate |
| gemma-4-26B-A4B-it-qat-q4_0-unquantized | UD-Q4_K_XL | 14.2 GB | 51.5 estimate |
| gemma-4-12B-it-qat-q4_0-unquantized | UD-Q4_K_XL | 6.7 GB | 33.6 estimate |
| gemma-3-1b-it | Q8_0 | 1.1 GB | 188 estimate |
Similar hardware
NVIDIA L40 · NVIDIA A40 · NVIDIA Tesla V100 32GB · NVIDIA A30 · NVIDIA Tesla P40 · NVIDIA Tesla M40 · NVIDIA L4 · NVIDIA A10