Prefill and token generation speed
Qwen Flash Next on two DGX Sparks: what changes?
Explore Qwen3.8 Flash Next on two DGX Sparks. Compare the same model on one unit and separate added model capacity from single-response generation speed.
Benchmark-based estimate · 4K input · one user
Last updated
Change device or model
Expand settings Settings4-bit · 4K · Baseline decodeCollapse
Reading this result
Two units are connected nodes, not one memory pool. Partitioning and interconnect conditions affect results, and settings for one fast request differ from those for throughput. This screen simulates a single request.
- Required / available memory
- 107.1GB / 232GB
- Runtime
- 4-bit · 2-node setup
Compare this model side by side
Compare from prefill against one unit. If a second unit does not speed up a single answer as much as you expect, reconsider whether your goal is larger models or concurrent throughput.
Check configuration
2× DGX Spark 256GB · Memory 256GB
Qwen3.8-Flash-Next Q4_K_M configuration. Check that the retailer lists the same memory option.
What changes with a longer document?Expand input length, waiting time, speed and memory
Detailed analysis and all 60 hardware configurations
Popular configurations
Speed simulations by device and model
Studio M3 Ultra 512GB × Qwen3.8-Flash-Next
50 ~ 70.3 tok/s · 3.3 to 4.5 seconds
Qwen Flash Next on Mac Studio M5 Ultra 512GB
103.5 ~ 184.1 tok/s · 0.65 to 1.2 seconds
RTX PRO 6000 96GB × Qwen3.8-Flash-Next
175.2 ~ 340.0 tok/s · 1.2 to 2.2 seconds
2× RTX 5090 64GB (PCIe) × Qwen3.8-Flash-Next
127.7 ~ 263.9 tok/s · 0.46 to 0.94 seconds