DeepSeek V4 Flash DSpark posts strong throughput on dual DGX Spark systems
sqrlmstr5000 · reddit · 2026-07-24
DeepSeek V4 Flash DSpark is getting strong throughput on dual DGX Spark boxes
A user says DeepSeek V4 Flash DSpark has become the fastest and best local-model setup they have tried on dual DGX Spark / Dell GB10 machines.
They point to a dashboard screenshot showing token throughput over time and claim:
- the experience in VS Code feels much faster than Copilot
- quality is roughly at the level of Sonnet 4.6 in their subjective view
- long-running tasks succeed about 99% of the time
- planning and instruction-following are strong
The evidence is mainly operational: a monitoring graph for model tokens/sec and prompt tokens/sec across a day, with peaks in the low-50s t/s for generation and over 4K t/s for prompt processing.
More from Infra
- Anthropic's S-1 Reveals Over $600B in Infrastructure Commitments Across Google, Amazon, Microsoft — cedric_chee · 2026-10-26
- Anthropic S-1 reveals 48 cents of compute cost per dollar of Q2 revenue — nmasc_ · 2026-10-26
- Persistent AI joins UR network to route private AI inference with encrypted, IP-hidden traffic — 0xSammy · 2026-10-26
- Anthropic's disclosed compute commitments: $114.5B Google, $161.2B Broadcom top the list — BenBajarin · 2026-10-26
- SpaceX strikes $40B deal with Apollo Global to lease AI chips and data center hardware — Polymarket · 2026-10-26
- SpaceX AI revenue up 830%, Claude designs 318K antivenom proteins in 64 hours — PeterDiamandis · 2026-10-26