Can Qwen3 32B run on the DGX Spark?
Yes. At 8K context the best fit is F16, using about 63.7 GB of the 109 GB usable on the DGX Spark. Estimated speed ceiling: ~4 tokens/sec.
Memory by quant and context
| Quant | 4K ctx | 8K ctx | 32K ctx | Max tok/s (est., 8K) |
|---|---|---|---|---|
| 2-bit (Q2_K) | 13.1 GB Fits | 14.1 GB Fits | 20.1 GB Fits | 22 |
| 3-bit (Q3_K_M) | 16.6 GB Fits | 17.6 GB Fits | 23.6 GB Fits | 17 |
| 4-bit (Q4_K_M) • | 20.1 GB Fits | 21.1 GB Fits | 27.1 GB Fits | 14 |
| 5-bit (Q5_K_M) • | 23.3 GB Fits | 24.3 GB Fits | 30.3 GB Fits | 12 |
| 6-bit (Q6_K) • | 26.7 GB Fits | 27.7 GB Fits | 33.7 GB Fits | 10 |
| 8-bit (Q8_0) • | 34.1 GB Fits | 35.1 GB Fits | 41.1 GB Fits | 8 |
| 16-bit (F16/BF16) | 62.7 GB Fits | 63.7 GB Fits | 69.7 GB Fits | 4 |
• real GGUF file size from Qwen/Qwen3-32B-GGUF. Speed is a bandwidth ceiling (273 GB/s), not a benchmark. Methodology.
What fits instead
- Cheapest listed GPU that runs Qwen3 32B: RTX 5090 (32 GB, launch MSRP $1,999) at Q5_K_M, fully in VRAM at 8K context.
- Largest popular model that fits the DGX Spark: gpt-oss 120B (117B, MoE) at F16, about 61.9 GB at 8K context.
"Runs" and "fits" mean the whole model in GPU memory at 8K context (4-bit or better for the model pick). Prices are US launch MSRPs, not today's street prices. "Popular" means the 60 most-downloaded models we track.
More
Qwen3 32B 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 (Qwen/Qwen3-32B), NVIDIA specs. Model data last checked 2026-11-17; refreshed hourly.