NVIDIA DGX Spark vs Framework Desktop 128GB
Prices, stock and measured speed side by side. Last checked 2026-10-16.
Verdict
Framework Desktop 128GB is cheaper in USD by $1,250 (36%). In EUR, Framework Desktop 128GB is cheaper by €287 (at least one EUR price is converted). On gpt-oss-120b, NVIDIA DGX Spark is faster (58.72 vs 52.74 tok/s; sources and engines differ, see below). Per dollar, Framework Desktop 128GB gives more large-model speed (15.3 vs 12.5 tok/s per $1k).
Generated from the numbers below; no hands-on testing.
| NVIDIA DGX Spark (Founders Edition) | Framework Desktop 128GB (DIY) | |
|---|---|---|
| Price (USD) | $4,699 | $3,449 |
| Price (EUR) | ≈€4,176 | €3,889 |
| Memory | 128GB | 128GB |
| Stock | not tracked shop | not tracked shop |
| gpt-oss-120b tok/s | 58.72 | 52.74 |
| gpt-oss-120b tok/s per $1k | 12.5 | 15.3 |
| gpt-oss-120b tok/s per €1k | 14.1 | 13.6 |
| Llama-2 7B tok/s | 57.21 | 50.95 |
| Llama-2 7B tok/s per $1k | 12.2 | 14.8 |
| Llama-2 7B tok/s per €1k | 13.7 | 13.1 |
| Stores | US store · EU store | US store · EU store |
Benchmark sources
- NVIDIA DGX Spark: large-model source (llama.cpp CUDA, llama-bench tg32, b7946) · small-model source
- Framework Desktop 128GB: large-model source (llama.cpp Vulkan, llama-bench tg32 (comment in thread)) · small-model source
Large: gpt-oss-120b MXFP4 (59 GiB weights), text generation tok/s at empty context, single request. Small: llama.cpp llama-bench, Llama-2 7B Q4_0, tg128 (text generation tok/s); best of FA on/off as posted.