Builds / 2x Nvidia GB10 (unified, DGX Spark)
2x DGX Spark: Qwen3.8 Flash-Next NVFP4 on vLLM, 54.4 tok/s (Mia's AI Lab recipe)
- GPU(s)
- 2x Nvidia GB10 (unified, DGX Spark)
- VRAM / unified memory (GB)
- 256
- CPU
- Nvidia GB10 Grace (Arm) per node
Amazon · eBay (used) · Newegg - Motherboard
- not stated
- RAM
- 128 GB LPDDR5X unified per Spark (256 GB total)
- PSU
- not stated
- Other
- 2 NVIDIA DGX Sparks, multi-node
- Reported total cost
- not stated
- Power
- not stated
Measured speed (generation tokens/sec)
| Model | Quant | Engine | tok/s | Prompt processing (tok/s) | Details |
|---|---|---|---|---|---|
| Qwen3.8 Flash-Next | NVFP4 | vLLM, 2 nodes, 1 stream | 54.4 | 2,962 @ 32k-token prompt est. TTFT @8k ≈ 2.8 s (calculated from measured prefill) | context: 262k · 1M with YaRN📏 Single-stream decode as published on Mia's AI Lab recipe card (Decode speed, one stream). Peak throughput 207 tok/s ×8 streams. |
Prompt processing = prefill speed (how fast the prompt is read) as reported by the source, with the prompt length it was measured at; blank = the source does not report it. est. TTFT @8k is calculated from the measured prefill speed (8,192 ÷ prefill tok/s), shown only when prefill was measured on a prompt of 512+ tokens. It is not a measurement and ignores load time and caching. measured TTFT is the source's own time-to-first-token. Why prefill matters for coding assistants.
Recipe and numbers by Mia's AI Lab (mia-ai.net), recipe last updated 2026-09-29; RigBase read the recipe card on 2026-10-27 and checked the decode figure against the recipe README. RigBase did not re-run it. Prefill is prompt-processing speed on a 32k prompt where listed.
https://mia-ai.net/models
Blank fields = not stated in the source. We never estimate build data; GPU street prices above are separately sourced and dated.
Quick answers: Can 2x DGX Spark (GB10) run Qwen3.8 Flash-Next?