Can deepseek moe 16b base run on the DGX Spark?
Yes. At 8K context the best fit is F16, using about 32.8 GB of the 109 GB usable on the DGX Spark. Estimated speed ceiling: ~48 tokens/sec.
Memory by quant and context
| Quant | 4K ctx | Max tok/s (est., 8K) |
|---|---|---|
| 2-bit (Q2_K) • | 7.7 GB Fits | 236 |
| 3-bit (Q3_K_M) • | 9.4 GB Fits | 186 |
| 4-bit (Q4_K_M) • | 11.6 GB Fits | 146 |
| 5-bit (Q5_K_M) • | 13.0 GB Fits | 128 |
| 6-bit (Q6_K) • | 15.1 GB Fits | 108 |
| 8-bit (Q8_0) • | 17.7 GB Fits | 91 |
| 16-bit (F16/BF16) | 32.0 GB Fits | 48 |
• real GGUF file size from mradermacher/deepseek-moe-16b-base-GGUF. Speed is a bandwidth ceiling (273 GB/s), not a benchmark. Methodology.
More
deepseek moe 16b base on every GPU and Mac · Every model that runs on the DGX Spark · Other models that fit 64 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.