How fast is Ornith-1.5-35B-A3B on Strix Halo?

The fastest of 5 measured configs is 105.6 tok/s single-stream (FP4, speculative decode, mtp). The slowest is 60 tok/s.

Which quant of Ornith-1.5-35B-A3B is fastest on Strix Halo?

FP4 at 76.9 tok/s, ahead of Q4_K_M at 71.6. Plain decoding only.

Configs › Ornith-1.5-35B-A3B › Ornith-1.5-35B-A3B on Strix Halo: 105.6 tok/s, 5 measured configs

Ornith-1.5-35B-A3B on Strix Halo: 105.6 tok/s, 5 measured configs

Measured Ornith-1.5-35B-A3B speeds on Strix Halo: fastest 105.6 tok/s single-stream (FP4, speculative decode, mtp), 5 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 Ornith-1.5-35B-A3B on Strix Halo: 105.6 tok/s (FP4, speculative decode, mtp), snapshot 2026-10-29.

Every measured single-stream run of Ornith-1.5-35B-A3B on Strix Halo

Decode tok/s	Quant	Backend	Trust	Measured by	Source checked	Source
105.6	FP4	speculative decode, mtp	Owner-submitted	julianmb	2026-09-10	source
76.9	FP4		Owner-submitted	julianmb	2026-09-16	source
75.9	Q4_K_M	llama.cpp/rocm (speculative decode, mtp-ngram)	Extracted, source-linked	c4software	2026-09-24	source
71.6	Q4_K_M		Owner-submitted	julianmb	2026-09-16	source
60	Q4_K_M	ollama	Extracted, source-linked	huppiflupp	2026-10-24	source

Best quant on Strix Halo

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

FP4: 76.9 tok/s, fastest ()
Q4_K_M: 71.6 tok/s, 7% lower ()

Questions and answers

How fast is Ornith-1.5-35B-A3B on Strix Halo?: The fastest of 5 measured configs is 105.6 tok/s single-stream (FP4, speculative decode, mtp). The slowest is 60 tok/s.
Which quant of Ornith-1.5-35B-A3B is fastest on Strix Halo?: FP4 at 76.9 tok/s, ahead of Q4_K_M at 71.6. Plain decoding only.

See also: Ornith-1.5-35B-A3B on every box · Strix Halo hardware page (from S$4,329 in Singapore today).

Last updated: snapshot 2026-10-29; newest source check 2026-10-24.

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