DGX Spark 128GB: 128GB VRAM — What Local LLMs Can It Run?
Check prices
The DGX Spark 128GB has 128GB of LPDDR5X memory with 273 GB/s of bandwidth. No used-price data yet. At Q4 quantization it runs models up to 70B parameters at roughly 5 tokens/s.
Full DGX Spark 128GB specifications
What LLMs can the DGX Spark 128GB run?
One reference model per family (its main version)| Model | Q4_K_M | Q8_0 | FP16 | ||||||
|---|---|---|---|---|---|---|---|---|---|
| Fits | Speed t/s | Context | Fits | Speed t/s | Context | Fits | Speed t/s | Context | |
| DeepSeekDeepSeek R1 Qwen3 8B | ✓ | 35.5 | 128K | ✓ | 20.3 | 128K | ✓ | 11.2 | 128K |
| GemmaGemma 4 31B | ✓ | 8.8 | 32K | ✓ | 5.2 | 32K | ✓ | 2.9 | 32K |
| GLMGLM 4.7 Flash | ✓ | 71.3 | 128K | ✓ | 46 | 128K | ✓ | 27.4 | 128K |
| gpt-ossgpt-oss 20B | ✓ | 63.6 | 128K | ✓ | 40.1 | 128K | ✓ | 23.4 | 128K |
| KimiKimi Linear 48B A3B | ✓ | 76.3 | 128K | ✓ | 48.1 | 128K | ✓ | 28.1 | 128K |
| LlamaLlama 3.1 8B | ✓ | 36.5 | 128K | ✓ | 20.9 | 128K | ✓ | 11.4 | 128K |
| MistralMistral 7B v0.3 | ✓ | 39.9 | 32K | ✓ | 22.9 | 32K | ✓ | 12.6 | 32K |
| QwenQwen3 8B | ✓ | 35.5 | 32K | ✓ | 20.3 | 32K | ✓ | 11.2 | 32K |
Fits when weights + KV cache + runtime overhead is at or below usable memory, single card. "Offload" on MoE models means it runs with expert weights in system RAM (only attention layers and KV cache stay in VRAM); that speed assumes 70 GB/s RAM bandwidth. The context column is the largest tier that fits.
DGX Spark 128GB used price history
Not enough history to draw a chart yet — price tracking just started.
Prices by platform
Buy links may carry affiliate tags; they do not change the price you pay.
FAQ
How much VRAM does the DGX Spark 128GB have?
The DGX Spark 128GB has 128GB of LPDDR5X memory with 273 GB/s of bandwidth.
Can the DGX Spark 128GB run a 70B model?
Yes. At Q4_K_M with an 8K context a 70B dense model needs about 43GB, which fits in the DGX Spark 128GB's 128GB.
How much does a used DGX Spark 128GB cost?
No used-price data yet; it will appear here once the feed is live.