Qwen3.8-9B

18 records across 6 hardware.

hardware	model	quant	backend	tok/s	pp tok/s	W	ttft s	ctx/tg	date	class	flags	source
M4 Pro	Qwen3.8-9B	MXFP4	oMLX	51.7	377.1	–	–	1024	2026-10-13	community	–	source
M4 Pro	Qwen3.8-9B	MXFP4	oMLX	51.2	376.9	–	–	1024	2026-10-13	community	–	source
M4 Pro	Qwen3.8-9B	MXFP4	oMLX	45.4	347.9	–	–	16384	2026-10-13	community	–	source
M1 Pro	Qwen3.8-9B	4bit	oMLX	30.7	186.9	–	–	4096	2026-10-11	community	–	source
M1 Pro	Qwen3.8-9B	4bit	oMLX	30.7	186.9	–	–	4096	2026-10-11	community	–	source
M6	Qwen3.8-9B	4bit	oMLX	30.0	1055.0	–	–	1024	2026-10-10	community	–	source
M6	Qwen3.8-9B	4bit	oMLX	29.4	1121.0	–	–	4096	2026-10-10	community	–	source
M1 Pro	Qwen3.8-9B	4bit	oMLX	27.6	162.0	–	–	16384	2026-10-11	community	–	source
M1 Pro	Qwen3.8-9B	4bit	oMLX	27.6	162.0	–	–	16384	2026-10-11	community	–	source
M1 Pro	Qwen3.8-9B	4bit	oMLX	26.4	158.1	–	–	1024	2026-10-18	community	–	source
M1 Pro	Qwen3.8-9B	4bit	oMLX	25.9	198.2	–	–	4096	2026-10-18	community	contradiction	source
M5	Qwen3.8-9B	4bit	oMLX	24.2	647.0	–	–	16384	2026-10-16	community	–	source
M5	Qwen3.8-9B	4bit	oMLX	21.7	582.6	–	–	32768	2026-10-16	community	–	source
M4	Qwen3.8-9B	4bit	oMLX	20.3	240.9	–	–	4096	2026-10-09	community	–	source
M4	Qwen3.8-9B	4bit	oMLX	18.2	229.8	–	–	16384	2026-10-09	community	–	source
M5	Qwen3.8-9B	4bit	oMLX	17.2	455.4	–	–	65536	2026-10-16	community	–	source
AMD Custom GPU 0932	Qwen3.8-9B 9.20B	Q4_K - Medium	llama.cpp (Vulkan)	12.22	101.16	–	–	tg128	2026-08-20	community	–	source
AMD Custom GPU 0932	Qwen3.8-9B 9.20B	Q4_K - Medium	llama.cpp (Vulkan)	12.08	100.16	–	–	tg128	2026-08-20	community	–	source