DeepSeek-V4-Flash hits 48.9 tok/s on single DGX Spark with speculative decoding
max_paperclips · x · 2026-08-03
Developer @danpacary successfully ran the 81GB DeepSeek-V4-Flash-0731 model on a single DGX Spark.
- Hardware: 128GB memory
- Plain decode: 22.5 tok/s
- Speculative decoding: Adding a 6.5GB drafter model boosted speed to 48.9 tok/s
The test utilizes the ds4 inference engine developed by antirez.
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