Can NVIDIA DGX Spark 64GB run Qwen3.6 35B-A3B?
Measured results for Qwen3.6 35B-A3B (35.95B MoE, 3B active) on NVIDIA DGX Spark 64GB, plus what fits and the commands to run it.
Short answer
Yes. Qwen3.6 35B-A3B Q8_0 fits fully on NVIDIA DGX Spark 64GB at 16k context, using 37.3 of 53.9 GB usable, and generates about 60 tokens/s. For Q8_0, our bandwidth-only estimate was ~49 tok/s against a measured median of ~49 (within 3%).
Measured generation speed by quantization
| Quant | Generation tok/s | Range | Prompt tok/s | Runs | Origin |
|---|---|---|---|---|---|
| Q4 class (median) | 60 | 60–60 | 2789 | 1 | public runs |
| Q8 class (median) | 49.2 | 49.2–49.2 | 2572 | 1 | public runs |
Generation = tg128 (128 generated tokens, short context). Public rows are medians of community llama.cpp results (sources); our own runs use llama-bench on a rented GPU.
Alternatives
Less memory that still fits Qwen3.6 35B-A3B
Smaller memory usually means a cheaper card or Mac (we do not list prices).
- AMD RX 7900 XTX 24GB: UD-Q4_K_M, 23 GB usable, ~249 tok/s estimated
- RTX 4090 24GB: UD-Q4_K_M, 23 GB usable, ~190 tok/s measured
- RTX 3090 24GB: UD-Q4_K_M, 23 GB usable, ~154 tok/s measured
Faster for Qwen3.6 35B-A3B
- RTX 5090 32GB: UD-Q6_K, 31 GB usable, ~260 tok/s measured
- 2x RTX 3090 (48GB total): Q8_0, 46 GB usable, ~165 tok/s estimated
- Mac Studio (2025) M3 Ultra 96GB: Q8_0, 67 GB usable, ~83 tok/s estimated
Other models on NVIDIA DGX Spark 64GB, similar size
- Ornith 1.5 35B-A3B (hardware page): Q8_0, fits fully, ~49 tok/s (scaled from measured)
- Laguna XS 2.1 (hardware page): Q8_0, fits fully, ~49 tok/s (scaled from measured)
- Nanbeige4.3 33B (hardware page): Q4_K_M, fits fully, ~11 tok/s (estimated)
- DeepSeek R1 Distill Qwen 32B (hardware page): Q4_K_M, fits fully, ~11 tok/s (estimated)
Memory and context limits
Q8_0 weights are 36.9 GB; the 16k f16 KV cache adds 0.40 GB. Largest context that fits in 53.9 GB usable:
| KV cache type | Max context |
|---|---|
| f16 | 512k |
| q8_0 | 1M |
| q4_0 | 1M |
Native limit 256k, 1.01M with YaRN. Full KV cache table for Qwen3.6 35B-A3B.
Copy-paste commands
The advisor's pick for this hardware: Q8_0, 16k context.
llama-server -hf unsloth/Qwen3.6-35B-A3B-GGUF:Q8_0 -c 16384 -fa on --jinja -ngl 99
OLLAMA_CONTEXT_LENGTH=16384 ollama serve ollama run hf.co/unsloth/Qwen3.6-35B-A3B-GGUF:Q8_0
FAQ
How fast is Qwen3.6 35B-A3B on NVIDIA DGX Spark 64GB?
About 60 tokens/s for generation (Q4 class, median of 1 public run). Speed drops as the context fills; see the depth table on this hardware's other pages.
How much memory does Qwen3.6 35B-A3B need on NVIDIA DGX Spark 64GB?
The advisor's pick, Q8_0, is 36.9 GB of weights plus 0.40 GB of KV cache at 16k context (37.3 GB), against 53.9 GB usable.
Is the Qwen3.6 35B-A3B speed on NVIDIA DGX Spark 64GB measured or estimated?
Measured: the median of 1 public llama.cpp result for the Q4 quant class.
What can I use instead of NVIDIA DGX Spark 64GB for Qwen3.6 35B-A3B?
With less memory, AMD RX 7900 XTX 24GB (~249 tok/s), RTX 4090 24GB (~190 tok/s), RTX 3090 24GB (~154 tok/s) also fit it fully. Faster: RTX 5090 32GB (~260 tok/s), 2x RTX 3090 (48GB total) (~165 tok/s), Mac Studio (2025) M3 Ultra 96GB (~83 tok/s).
What is the maximum context for Qwen3.6 35B-A3B on NVIDIA DGX Spark 64GB?
With an f16 KV cache about 512k tokens, 1M with q8_0 and 1M with q4_0, bounded by the model's own limit.
Related
All models on NVIDIA DGX Spark 64GB · Qwen3.6 35B-A3B on all hardware · VRAM tier · all measured pages · leaderboard