Configs › DeepSeek-V4-Flash-0731 tok/s: Strix Halo 35.7

DeepSeek-V4-Flash-0731 tok/s: Strix Halo 35.7

Measured DeepSeek-V4-Flash-0731 speeds: 35.7 tok/s on Strix Halo; 32.8 tok/s on DGX Spark (vllm). 15 configs, each linked to its source, snapshot 2026-10-16.

Snapshot 2026-10-16 · 394 configs on AI boxes and 423 on GPU cards · 123 models. Every number is generated from the tracker snapshot and linked to where it was measured; nothing here is typed by hand.

Quick answers

How fast is DeepSeek-V4-Flash-0731 on Strix Halo?: 35.7 tok/s single-stream decode (UD-IQ3_XXS, llama.cpp/vulkan (speculative decode, dspark)), the fastest of 14 measured configs on AMD Strix Halo (Ryzen AI Max+ 395), measured by pepuscz, source last checked 2026-10-16. source
How fast is DeepSeek-V4-Flash-0731 on DGX Spark?: 32.8 tok/s single-stream decode (vllm), the only measured config on NVIDIA DGX Spark (GB10), measured by jvr0x, source last checked 2026-10-14. source

Measured configs for DeepSeek-V4-Flash-0731

Model	Hardware	Quant	Backend	Decode	Context	Trust	Source
DeepSeek-V4-Flash-0731	Strix Halo	UD-IQ3_XXS	llama.cpp/vulkan (speculative decode, dspark)	35.7 tok/s (prefill 226.8)	524288	Structured table	source
DeepSeek-V4-Flash-0731	Strix Halo	2.58bpw-mix	speculative decode, dspark	34.2 tok/s		Extracted, source-linked	source
DeepSeek-V4-Flash-0731	DGX Spark		vllm	32.8 tok/s		Structured table	source
DeepSeek-V4-Flash-0731	Strix Halo		rocm (speculative decode, dspark)	29.6 tok/s (prefill 142.8)	131072	Structured table	source
DeepSeek-V4-Flash-0731	Strix Halo	UD-IQ3_XXS	llama.cpp/vulkan-radv-performance	19.8 tok/s		Structured table	source
DeepSeek-V4-Flash-0731	Strix Halo	UD-IQ2_XXS	llama.cpp/rocm-7.14	16.2 tok/s		Structured table	source
DeepSeek-V4-Flash-0731	Strix Halo	UD-IQ2_XXS	llama.cpp/rocm-7.2.4	16.1 tok/s		Structured table	source
DeepSeek-V4-Flash-0731	Strix Halo	UD-IQ3_XXS	llama.cpp/rocm-7.14	16 tok/s		Structured table	source
DeepSeek-V4-Flash-0731	Strix Halo	UD-IQ3_XXS	llama.cpp/rocm-7.2.4	15.8 tok/s		Structured table	source
DeepSeek-V4-Flash-0731	Strix Halo	UD-IQ2_XXS	llama.cpp/rocm-7.14-pr26592	14.8 tok/s		Structured table	source
DeepSeek-V4-Flash-0731	Strix Halo	UD-IQ3_XXS	llama.cpp/rocm-7.14-pr26592	14.5 tok/s		Structured table	source
DeepSeek-V4-Flash-0731	Strix Halo	UD-IQ2_XXS	llama.cpp/vulkan-radv-performance	13.1 tok/s		Structured table	source
DeepSeek-V4-Flash-0731	Strix Halo	UD-IQ2_XXS	vulkan	12 tok/s (prefill 130)	131072	Verified (ran it)	source
DeepSeek-V4-Flash-0731	Strix Halo	UD-IQ2_XXS	llama.cpp/vulkan-radv	9.1 tok/s		Structured table	source
DeepSeek-V4-Flash-0731	Strix Halo	UD-IQ3_XXS	llama.cpp/vulkan-radv	9 tok/s		Structured table	source

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