DGX Spark (GB10): Qwen3.6 35B-A3B NVFP4 at 113 tok/s on vLLM with MTP
| GPU(s) | Nvidia GB10 (unified) |
|---|---|
| VRAM / unified memory (GB) | 128 |
| CPU | Nvidia GB10 (20 Arm cores) Amazon · eBay (used) · Newegg |
| Motherboard | |
| RAM | 128 GB unified |
| PSU | |
| Other | NVIDIA DGX Spark; vLLM 0.30, NVFP4 weights, MTP speculative decoding |
| Reported total cost | $3,999 (≈€3,568) |
| Power |
Measured speed (generation tokens/sec)
📏 How measured: reported by the original source; RigBase re-checked the numbers at that source
| Model | Quant | Engine | tok/s | Details |
|---|---|---|---|---|
| Qwen3.6 35B-A3B | NVFP4 | vLLM 0.30 + MTP | 113 | ⚡ spec-decode (MTP) |
| Qwen3.6 35B-A3B | NVFP4 | vLLM 0.30 + MTP | 86 | ⚡ spec-decode (MTP) |
Measured 6 Oct 2026, median of 4 runs. 113 tok/s = thinking-mode workload; 86 = multi-step assistant session; 82 on short structured JSON. Prefill: 100K-token cold prompt in about 30 s. 16 concurrent requests: 311 tok/s total (23 per request). Cost: author says he paid about $3,999 at launch.
Source: https://vitorpontual.com/posts/two-ai-boxes-3-the-benchmarks/
Blank fields = not stated in the source. We never estimate build data; GPU street prices above are separately sourced and dated.
Rent before you buy: try this model on a cloud GPU for a few dollars first: Runpod (referral link; sign up with Google to get the new-user bonus)
Quick answers: Can DGX Spark (GB10) run Qwen3.6 35B-A3B?