Can an NVIDIA DGX Spark run DeepSeek-R1?
No. DeepSeek-R1 needs 405.6GB at Q4_K_M with llama.cpp and 8K context; the NVIDIA DGX Spark has 128GB of unified memory (~90% GPU-addressable) (114.7GB usable) — it does not fit, even with CPU offload.
NVIDIA DGX Spark · DeepSeek-R1 · Q4_K_M
❌ Won't fit
betaMLA memory formula — beta, not yet calibrated · New hardware class — not yet calibrated
weights 404.4 GBKV cache 0.58 GBoverhead 0.58 GBpool 114.7 GBneeds 405.6 GB — over the pool
assumes llama.cpp · batch 1 · 8K context · fp16 KV cache
No quant of this model fits cleanly on this hardware with llama-cpp at this context — a smaller model or more memory is the real fix.
$4,699 MSRP · as of 2026-07-11
Affiliate link — we may earn a commission. Verdicts are computed before any link is attached.
Every quant, graded
NVIDIA DGX Spark × DeepSeek-R1, llama.cpp, 8K context.
| Quant | File size | Fit | Speed tier |
|---|---|---|---|
| Q8_0 | 713.3 GB | Won't fit | |
| Q6_K | 550.8 GB | Won't fit | |
| Q5_K_M | 475.4 GB | Won't fit | |
| Q4_K_M | 404.4 GB | Won't fit |
DeepSeek-R1 on similar hardware
More models on the NVIDIA DGX Spark
Verdict computed from published specs and measured GGUF file sizes — see how the math works.