Builds / Nvidia GB10 (unified)
DGX Spark (GB10, 128GB): Nemotron 3.5 Super VL 120B and Bolt 74B, launch-week runs
Source: dev.classmethod.jp ↗More source: dev.classmethod.jp ↗More source: forums.developer.nvidia.com ↗📏 How measured: reported by the original source; RigBase re-checked the numbers at that source
- GPU(s)
- Nvidia GB10 (unified)
- VRAM / unified memory (GB)
- 128
- CPU
- Nvidia GB10
Amazon · eBay (used) · Newegg - Motherboard
- not stated
- RAM
- 128GB unified memory
- PSU
- not stated
- Other
- not stated
- Reported total cost
- not stated
- Power
- GPU peak 82 W (nvidia-smi) during Super VL runs with prompts up to 64K tokens
Measured speed (generation tokens/sec)
| Model | Quant | Engine | tok/s | Details |
|---|---|---|---|---|
| Nemotron 3.5 Super VL 120B-A12B | NVFP4, no speculative decoding, 256-token output | vLLM | 19.06 | |
| Nemotron 3.5 Super VL 120B-A12B | NVFP4 + LiLiCorr speculative decoding (7 draft tokens) | vLLM | 38.07 | ⚡ spec-decode (LiLiCorr) |
| Nemotron 3.5 Bolt 74B-A9B | 2-bit 24GB GGUF, no speculative decoding | llama.cpp | 39.47 | |
| Nemotron 3.5 Bolt 74B-A9B | 2-bit 24GB GGUF + DFlash speculative decoding (7 draft tokens) | llama.cpp | 61.61 | ⚡ spec-decode (DFlash) |
DevelopersIO (Classmethod) articles of 16 and 18 Oct 2026, measured on the author's own DGX Spark with the model cards' Spark recipes; median of 3 runs, temperature 0. Super VL ran in vLLM (nightly image from the model card), Bolt in llama.cpp. Speculative-decoding speeds depend on how predictable the text is. The NVIDIA forum thread (third link) reports hard power-offs on some Sparks during long (32K+) Super VL prompts.
Source: https://dev.classmethod.jp/articles/dgx-spark-nemotron-3-5-super-vl-first-touch/
https://dev.classmethod.jp/articles/dgx-spark-nemotron-3-5-bolt-74b-a9b-first-touch/
https://forums.developer.nvidia.com/t/nemotron-3-5-super-vl-120b-a12b-nvfp4-on-a-single-dgx-spark-34-49-tok-s-decode-with-lilicorr-16k-thermal-shutdowns-at-32k/385991
https://dev.classmethod.jp/articles/dgx-spark-nemotron-3-5-bolt-74b-a9b-first-touch/
https://forums.developer.nvidia.com/t/nemotron-3-5-super-vl-120b-a12b-nvfp4-on-a-single-dgx-spark-34-49-tok-s-decode-with-lilicorr-16k-thermal-shutdowns-at-32k/385991
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 Nemotron 3.5 Bolt 74B-A9B? · Can DGX Spark (GB10) run Nemotron 3.5 Super VL 120B-A12B?