Lucebox Beats NVIDIA DGX Spark by 3.6x in DeepSeek Decode
samsja19 · x · 2026-08-03
Lucebox partnered with AMD to outperform the NVIDIA DGX Spark in DeepSeek V4 Flash decode speed using heterogeneous consumer hardware.
Performance & Specs
- Achieves 51.1 tok/s on the full 284B model, 3.63x faster than a single DGX Spark (14.09 tok/s).
- Hardware: AMD Radeon AI PRO R9700 (dense path/hot experts) + Strix Halo (128GB unified memory for parallel expert computation).
- Priced at $6,499 ($2,899 less than two DGX Sparks), focusing on plug-and-play private local inference.
Related event: DeepSeek V4 Flash Sparks a Wave of Local Deployment Tests(26 posts)→
More from Infra
- NVIDIA quietly delists the entry-level RTX 5050 from its marketplace, following the RTX 5090 — chemist_slime · 2026-10-16
- Grok claims 816B MoE model runs locally on dual M5 Ultra Mac Studios via MLX — grok · 2026-10-16
- Same 1,000 Cloudflare Durable Objects: $4,050/month cut to $22 with one trick — threepointone · 2026-10-16
- NVIDIA ships FP4-quantized Qwen3.8-Flash for local use, trending on Hugging Face — nvidia · 2026-10-16
- Baidu's Virtual Head Attention tames the KV cache, FLOPs, quality triangle in long-context LLMs — baidu · 2026-10-16
- DuoDiscover agents auto-build fast eval proxies, scoring 2.68x TPU kernel speedup — berkeley_ai · 2026-10-16