How fast is Qwen3.8-Flash-Next on DGX Spark?

The fastest of 8 measured configs is 44.2 tok/s single-stream (NVFP4, vllm (speculative decode, mtp-2)). The slowest is 22.3 tok/s.

Which backend is fastest for Qwen3.8-Flash-Next on DGX Spark?

atlas at 36.7 tok/s, then vllm at 35.1. Each backend figure is its best plain-decoding row and may use a different quant.

How does DGX Spark compare with other boxes for Qwen3.8-Flash-Next?

DGX Spark best 44.2 tok/s. Other boxes, best measured: Mac Ultra 76.4, Strix Halo 63.7, Mac Max 57.5. Backends and quants differ, so this is the best known number per box, not a controlled test.

Configs › Qwen3.8-Flash-Next › Qwen3.8-Flash-Next on DGX Spark: 44.2 tok/s, 8 measured configs

Qwen3.8-Flash-Next on DGX Spark: 44.2 tok/s, 8 measured configs

Measured Qwen3.8-Flash-Next speeds on DGX Spark: fastest 44.2 tok/s single-stream (NVFP4, vllm (speculative decode, mtp-2)), 8 configs with quant, backend and context, each linked to its source, snapshot 2026-10-29.

Snapshot 2026-10-29 · 413 configs on AI boxes and 288 on GPU cards · 121 models. Every number is generated from the tracker snapshot and linked to where it was measured; nothing here is typed by hand.

Fastest measured single-stream run of Qwen3.8-Flash-Next on DGX Spark: 44.2 tok/s (NVFP4, vllm (speculative decode, mtp-2)), snapshot 2026-10-29.

Every measured single-stream run of Qwen3.8-Flash-Next on DGX Spark

Decode tok/s	Quant	Backend	Context	Trust	Measured by	Source checked	Source
44.2	NVFP4	vllm (speculative decode, mtp-2)	262144	Extracted, source-linked	YSLAB-ai	2026-09-01	source
39.8	NVFP4	vllm (speculative decode, mtp-3)		Extracted, source-linked	codegyver	2026-09-06	source
36.7	NVFP4	atlas		Extracted, source-linked	Atlas-Inf	2026-09-23	source
35.1	NVFP4	vllm		Structured table	jvr0x	2026-10-29	source
27.6	NVFP4	vllm (speculative decode, mtp-1)		Extracted, source-linked	sayyidfareed	2026-08-29	source
27.3	NVFP4	vllm (baseline)	262144	Extracted, source-linked	YSLAB-ai	2026-08-29	source
26.7	NVFP4	vllm	1000000	Extracted, source-linked	sayyidfareed	2026-08-29	source
22.3	NVFP4			Extracted, source-linked	Soulmate-Halo	2026-10-14	source

Backends on DGX Spark

Fastest plain-decoding run per backend. Quants can differ between rows.

atlas: 36.7 tok/s (NVFP4)
vllm: 35.1 tok/s (NVFP4)

Qwen3.8-Flash-Next on other boxes

Mac Ultra: 76.4 tok/s best of 1 config
Strix Halo: 63.7 tok/s best of 15 configs
Mac Max: 57.5 tok/s best of 2 configs

Questions and answers

How fast is Qwen3.8-Flash-Next on DGX Spark?: The fastest of 8 measured configs is 44.2 tok/s single-stream (NVFP4, vllm (speculative decode, mtp-2)). The slowest is 22.3 tok/s.
Which backend is fastest for Qwen3.8-Flash-Next on DGX Spark?: atlas at 36.7 tok/s, then vllm at 35.1. Each backend figure is its best plain-decoding row and may use a different quant.
How does DGX Spark compare with other boxes for Qwen3.8-Flash-Next?: DGX Spark best 44.2 tok/s. Other boxes, best measured: Mac Ultra 76.4, Strix Halo 63.7, Mac Max 57.5. Backends and quants differ, so this is the best known number per box, not a controlled test.

See also: Qwen3.8-Flash-Next on every box · DGX Spark hardware page (from S$7,452 in Singapore today).

Last updated: snapshot 2026-10-29; newest source check 2026-10-29.

More: all configs · hardware · news · contributors · methodology · llms.txt · llms-full.txt · API (OpenAPI) · JSON snapshot. Built by Altronis, private on-prem AI, Singapore.