Can deepseek moe 16b base run on the L40S?
Yes. At 8K context the best fit is F16, using about 32.8 GB of the 45.6 GB usable on the L40S. Estimated speed ceiling: ~153 tokens/sec.
Memory by quant and context
| Quant | 4K ctx | Max tok/s (est., 8K) |
|---|---|---|
| 2-bit (Q2_K) • | 7.7 GB Fits | 746 |
| 3-bit (Q3_K_M) • | 9.4 GB Fits | 590 |
| 4-bit (Q4_K_M) • | 11.6 GB Fits | 461 |
| 5-bit (Q5_K_M) • | 13.0 GB Fits | 404 |
| 6-bit (Q6_K) • | 15.1 GB Fits | 341 |
| 8-bit (Q8_0) • | 17.7 GB Fits | 287 |
| 16-bit (F16/BF16) | 32.0 GB Fits | 153 |
• real GGUF file size from mradermacher/deepseek-moe-16b-base-GGUF. Speed is a bandwidth ceiling (864 GB/s), not a benchmark. Methodology.
More
deepseek moe 16b base on every GPU and Mac · Every model that runs on the L40S · Other models that fit 48 GB VRAM · Try other settings in the calculator
Data: Hugging Face Hub (deepseek-ai/deepseek-moe-16b-base), NVIDIA specs. Data as of 2026-11-17.