Best Local LLMs for the NVIDIA DGX Spark
Autor: Jakub Rusinowski · Ostatnia aktualizacja: 21 lipca 2026
Best all-round pick: GPT-OSS 120B
The NVIDIA DGX Spark has 128 GB of VRAM, of which about 128 GB is available to a model. The largest model it can hold is Step 3.7 Flash 198B-A11B (196B, 112.3 GB at Q4_K_M).
Best models overall on the NVIDIA DGX Spark
| Model | Score | Memory at Q4_K_M | Context | Licence |
|---|---|---|---|---|
| GPT-OSS 120B | 95.8 | 71.3 GB | 128K | Apache-2.0 |
| Qwen 3.5 122B-A10B (MoE) | 94.2 | 74.5 GB | 125K | Apache 2.0 |
| Qwen 3.5 122B-A10B | 92.5 | 74.5 GB | 128K | Apache-2.0 |
| GLM-5.1 72B | 91.2 | 44.3 GB | 125K | MIT |
| Command R+ (104B) | 91.2 | 63.6 GB | 125K | CC-BY-NC |
| Llama 4 Scout 17B | 91.2 | 66.6 GB | 10M | Llama Community |
Best model by what you are doing
| Workload | Recommended on this GPU |
|---|---|
| Coding | Devstral-2 123B (99) Qwen3-Coder-Next (80B-A3B MoE) (96.6) Qwen 3.5 122B-A10B (MoE) (96.5) |
| General assistant | GPT-OSS 120B (95.8) Qwen 3.5 122B-A10B (MoE) (94.2) Qwen 3.5 122B-A10B (92.5) |
| Reasoning | GLM-5.1 72B (98.4) Qwen 3.5 122B-A10B (MoE) (97) Gemma 4 27B ⭐ (95.9) |
| RAG | Llama 4 Scout 17B (93.6) Qwen3.8-Flash-Next (93) Gemma 4 31B (92.8) |
| Agents | Devstral-2 123B (95.2) Qwen3.8-Flash-Next (93.9) Mistral Small 4 119B-A6.5B (93.5) |
| Vision | Qwen 3.5 72B (96.9) Llama 3.2 90B Vision Instruct (95.2) Qwen3.8-Flash-Next (93.7) |
How these numbers are calculated
- Usable memory is 128 GB of the card's 128 GB.
- Fit is measured at Q4_K_M — quantized weights plus KV cache plus 0.8 GB runtime overhead.
- Within this card's budget, models are ranked on how well they use the memory available, not on being small — so the recommendation scales with the hardware.
FAQ
What is the best LLM for the NVIDIA DGX Spark?
GPT-OSS 120B is the strongest all-round pick that fits its 128 GB.
What is the largest model the NVIDIA DGX Spark can run?
Step 3.7 Flash 198B-A11B — 196B parameters, needing 112.3 GB at Q4_K_M.
How much of the NVIDIA DGX Spark's memory can a model actually use?
About 128 GB of its 128 GB, before the desktop and runtime overhead are accounted for.
Compatibility Checks for the NVIDIA DGX Spark
- Can I run Command R Family on the NVIDIA DGX Spark?
- Can I run Llama 4 on the NVIDIA DGX Spark?
- Can I run Llama 3.2 Family on the NVIDIA DGX Spark?
- Can I run Llama 3.2 Vision on the NVIDIA DGX Spark?
- Can I run Qwen 3.5 on the NVIDIA DGX Spark?
Similar GPUs
By Workload
- Best local LLMs for coding
- Best local LLMs for general assistant
- Best local LLMs for reasoning
- Best local LLMs for rag