How fast is Qwen3-Coder 30B-A3B on Strix Halo?

The fastest of 4 measured configs is 98 tok/s single-stream (Q4_K_S, llama.cpp/vulkan). The slowest is 73.7 tok/s.

Which quant of Qwen3-Coder 30B-A3B is fastest on Strix Halo?

Q4_K_S at 98 tok/s, ahead of UD-Q4_K_XL at 97.1, IQ4_XS at 90.4. Plain decoding only.

Which backend is fastest for Qwen3-Coder 30B-A3B on Strix Halo?

llama.cpp/vulkan at 98 tok/s, then llama.cpp/rocm at 73.7. Each backend figure is its best plain-decoding row and may use a different quant.

Configs › Qwen3-Coder 30B-A3B › Qwen3-Coder 30B-A3B on Strix Halo: 98 tok/s, 4 measured configs

Qwen3-Coder 30B-A3B on Strix Halo: 98 tok/s, 4 measured configs

Measured Qwen3-Coder 30B-A3B speeds on Strix Halo: fastest 98 tok/s single-stream (Q4_K_S, llama.cpp/vulkan), 4 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-Coder 30B-A3B on Strix Halo: 98 tok/s (Q4_K_S, llama.cpp/vulkan), snapshot 2026-10-29.

Every measured single-stream run of Qwen3-Coder 30B-A3B on Strix Halo

Decode tok/s	Quant	Backend	Context	Trust	Measured by	Source checked	Source
98	Q4_K_S	llama.cpp/vulkan	512	Structured table	hogeheer	2026-10-29	source
97.1	UD-Q4_K_XL	llama.cpp/vulkan	512	Structured table	hogeheer	2026-10-29	source
90.4	IQ4_XS	llama.cpp/vulkan	512	Structured table	hogeheer	2026-10-29	source
73.7	UD-Q4_K_XL	llama.cpp/rocm	512	Structured table	hogeheer	2026-10-29	source

Best quant on Strix Halo

Fastest plain-decoding run per quant, versus the fastest quant.

Q4_K_S: 98 tok/s, fastest (llama.cpp/vulkan)
UD-Q4_K_XL: 97.1 tok/s, 1% lower (llama.cpp/vulkan)
IQ4_XS: 90.4 tok/s, 8% lower (llama.cpp/vulkan)

Backends on Strix Halo

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

llama.cpp/vulkan: 98 tok/s (Q4_K_S)
llama.cpp/rocm: 73.7 tok/s (UD-Q4_K_XL)

Questions and answers

How fast is Qwen3-Coder 30B-A3B on Strix Halo?: The fastest of 4 measured configs is 98 tok/s single-stream (Q4_K_S, llama.cpp/vulkan). The slowest is 73.7 tok/s.
Which quant of Qwen3-Coder 30B-A3B is fastest on Strix Halo?: Q4_K_S at 98 tok/s, ahead of UD-Q4_K_XL at 97.1, IQ4_XS at 90.4. Plain decoding only.
Which backend is fastest for Qwen3-Coder 30B-A3B on Strix Halo?: llama.cpp/vulkan at 98 tok/s, then llama.cpp/rocm at 73.7. Each backend figure is its best plain-decoding row and may use a different quant.

See also: Qwen3-Coder 30B-A3B on every box · Strix Halo hardware page (from S$4,329 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.