NVIDIA · Blackwell · unified memory

What runs on a NVIDIA DGX Spark (GB10)

128 GB at 273 GB/s. The largest model that fits at Q4_K_M is gpt-oss 120B-A5.1B, at roughly 45.6 tokens per second. Unified memory means the operating system takes a share, so budget around three quarters of the total.

Huge capacity, modest bandwidth. Great for big MoE models, slow for big dense ones.

Runs comfortably

6.1 GB of 96.0 GB · 6%
096 GB
Weights 4.6 GB
KV cache 1.0 GB
Runtime overhead 0.5 GB

Llama 3.1 8B at Q4_K_M leaves 89.9 GB spare on a NVIDIA DGX Spark (GB10). There is room to raise the context length or move up a quantisation level.

Generation35.1tok/s
Prompt processing3269tok/s
Max context128Ktokens
KV per 1K tokens0GB

Every quantisation of Llama 3.1 8B on a NVIDIA DGX Spark (GB10)

Highlighted row is the highest quality that still fits at 8K context.

QuantisationbpwWeights TotalVerdictMax ctxtok/s
FP16 / BF16 16.00 15.0 GB 16.5 GB Runs comfortably 128K 11.6
INT8 / W8A8 8.50 7.7 GB 9.2 GB Runs comfortably 128K 21.9
Q8_0 (GGUF) 8.50 7.7 GB 9.2 GB Runs comfortably 128K 21.9
FP8 (E4M3) 8.00 7.5 GB 9.0 GB Runs comfortably 128K 22.5
Q6_K 6.56 6.1 GB 7.7 GB Runs comfortably 128K 27.1
AWQ 4-bit 4.25 5.4 GB 6.9 GB Runs comfortably 128K 30.4
GPTQ 4-bit 4.25 5.4 GB 6.9 GB Runs comfortably 128K 30.4
MXFP4 4.25 5.4 GB 6.9 GB Runs comfortably 128K 30.4
Q5_K_M 5.67 5.3 GB 6.8 GB Runs comfortably 128K 30.9
Q5_K_S 5.52 5.2 GB 6.7 GB Runs comfortably 128K 31.7
Q4_K_M 4.85 4.6 GB 6.1 GB Runs comfortably 128K 35.1
Q4_K_S 4.58 4.4 GB 5.9 GB Runs comfortably 128K 36.6
Q4_0 4.55 4.4 GB 5.9 GB Runs comfortably 128K 36.8
IQ4_XS 4.25 4.1 GB 5.6 GB Runs comfortably 128K 38.7
Q3_K_M 3.91 3.8 GB 5.4 GB Runs comfortably 128K 41.2
IQ3_M 3.70 3.7 GB 5.2 GB Runs comfortably 128K 42.8
IQ3_XXS 3.06 3.2 GB 4.7 GB Runs comfortably 128K 48.9
Q2_K 2.63 2.8 GB 4.3 GB Runs comfortably 128K 54.0
IQ2_XXS 2.06 2.3 GB 3.9 GB Runs comfortably 128K 62.6
IQ1_M 1.75 2.1 GB 3.6 GB Runs comfortably 128K 68.6

Models on a NVIDIA DGX Spark (GB10)

Q4_K_M at 8K context. Bandwidth sets the speed; capacity sets the ceiling.

ModelParamsWeightsVerdictMax ctxtok/s
DeepSeek-R1 671B-A37B 671B 379.0 GB Won't fit — 2.40
Qwen3 235B-A22B 235.1B 132.8 GB Won't fit — 6.20
gpt-oss 120B-A5.1B 116.8B 66.0 GB Runs comfortably 128K 45.6
Llama 4 Scout 109B-A17B 109B 61.7 GB Runs comfortably 180K 14.8
Qwen2.5 72B 72.7B 41.2 GB Runs comfortably 128K 4.24
Llama 3.1 70B 70.5B 40.0 GB Runs comfortably 128K 4.37
Mixtral 8x7B 46.7B 26.4 GB Runs comfortably 32K 19.7
Command R 35B 35.0B 20.1 GB Runs comfortably 60K 7.19
Yi-1.5 34B 34.4B 19.5 GB Runs comfortably 32K 8.80
Qwen3 32B 32.8B 18.6 GB Runs comfortably 128K 9.17
Qwen2.5-Coder 32B 32.8B 18.6 GB Runs comfortably 128K 9.17
DeepSeek-R1-Distill-Qwen 32B 32.8B 18.6 GB Runs comfortably 128K 9.17
Qwen3 30B-A3B 30.5B 17.3 GB Runs comfortably 128K 53.5
Gemma 3 27B 27.4B 15.6 GB Runs comfortably 128K 11.0
Mistral Small 24B 23.6B 13.4 GB Runs comfortably 32K 12.8
gpt-oss 20B-A3.6B 20.9B 11.9 GB Runs comfortably 128K 64.8
Qwen3 14B 14.8B 8.5 GB Runs comfortably 128K 19.8
Phi-4 14B 14.7B 8.4 GB Runs comfortably 16K 19.6
Mistral NeMo 12B 12.3B 7.0 GB Runs comfortably 128K 23.5
Gemma 3 12B 12.2B 7.0 GB Runs comfortably 128K 23.8
GLM-4 9B 9.4B 5.4 GB Runs comfortably 128K 32.2
Qwen3 8B 8.2B 4.7 GB Runs comfortably 128K 33.9
Llama 3.1 8B 8.0B 4.6 GB Runs comfortably 128K 35.1
DeepSeek-R1-Distill-Llama 8B 8.0B 4.6 GB Runs comfortably 128K 35.1
Qwen2.5 7B 7.6B 4.4 GB Runs comfortably 128K 38.9
Mistral 7B v0.3 7.3B 4.1 GB Runs comfortably 32K 38.8
Gemma 3 4B 4.3B 2.5 GB Runs comfortably 128K 66.4
Qwen3 4B 4.0B 2.3 GB Runs comfortably 128K 62.2
Llama 3.2 3B 3.2B 1.8 GB Runs comfortably 128K 78.1
Llama 3.2 1B 1.2B 0.7 GB Runs comfortably 128K 209

Compared with

NVIDIA DGX Spark (GB10) vs RTX 5090
128 GB vs 32 GBNVIDIA DGX Spark (GB10) vs RTX 4090
128 GB vs 24 GBNVIDIA DGX Spark (GB10) vs RTX 3090
128 GB vs 24 GBNVIDIA DGX Spark (GB10) vs RTX 5080
128 GB vs 16 GBNVIDIA DGX Spark (GB10) vs RTX 5070 Ti
128 GB vs 16 GBNVIDIA DGX Spark (GB10) vs RTX 4070 Ti Super
128 GB vs 16 GBNVIDIA DGX Spark (GB10) vs RTX 4060 Ti 16GB
128 GB vs 16 GBNVIDIA DGX Spark (GB10) vs RTX 3060 12GB
128 GB vs 12 GB

Direct answers

Llama 3.1 8B on NVIDIA DGX Spark (GB10)
8.0BLlama 3.1 70B on NVIDIA DGX Spark (GB10)
70.5BLlama 3.2 3B on NVIDIA DGX Spark (GB10)
3.2BLlama 3.2 1B on NVIDIA DGX Spark (GB10)
1.2BLlama 4 Scout 109B-A17B on NVIDIA DGX Spark (GB10)
109BQwen3 8B on NVIDIA DGX Spark (GB10)
8.2BQwen3 14B on NVIDIA DGX Spark (GB10)
14.8BQwen3 32B on NVIDIA DGX Spark (GB10)
32.8B