MI350X

18 records across 2 models (quants: MXFP4, unspecified).

hardware	model	quant	backend	tok/s	pp tok/s	W	ttft s	ctx/tg	date	class	flags	source
MI350X	DeepSeek-V4.1-Flash ?	unspecified	vLLM	1763	32443	–	–	16384	2026-10-05	sourced	–	source
MI350X	DeepSeek-V4.1-Flash ?	unspecified	vLLM	1717	31440	–	–	16384	2026-10-05	sourced	–	source
MI350X	MiniMax-M3	MXFP4	vLLM	910	–	–	–	8192	2026-10-01	community	–	source
MI350X	MiniMax-M3	MXFP4	vLLM	825	–	–	–	8192	2026-10-01	community	–	source
MI350X	MiniMax-M3	MXFP4	vLLM	682	–	–	–	8192	2026-10-01	community	–	source
MI350X	MiniMax-M3	MXFP4	vLLM	599	–	–	–	8192	2026-10-01	community	–	source
MI350X	MiniMax-M3	MXFP4	vLLM	574	–	–	–	8192	2026-10-01	community	–	source
MI350X	MiniMax-M3	MXFP4	vLLM	415	–	–	–	8192	2026-10-01	community	–	source
MI350X	MiniMax-M3	MXFP4	vLLM	405	–	–	–	8192	2026-10-01	community	–	source
MI350X	MiniMax-M3	MXFP4	vLLM	273	–	–	–	131072	2026-10-01	community	–	source
MI350X	MiniMax-M3	MXFP4	vLLM	254	–	–	–	8192	2026-10-01	community	–	source
MI350X	MiniMax-M3	MXFP4	vLLM	254	–	–	–	8192	2026-10-01	community	–	source
MI350X	MiniMax-M3	MXFP4	vLLM	246	–	–	–	131072	2026-10-01	community	–	source
MI350X	MiniMax-M3	MXFP4	vLLM	243	–	–	–	131072	2026-10-01	community	–	source
MI350X	MiniMax-M3	MXFP4	vLLM	206	–	–	–	131072	2026-10-01	community	–	source
MI350X	MiniMax-M3	MXFP4	vLLM	195	–	–	–	131072	2026-10-01	community	–	source
MI350X	MiniMax-M3	MXFP4	vLLM	157	–	–	–	131072	2026-10-01	community	–	source
MI350X	MiniMax-M3	MXFP4	vLLM	141	–	–	–	8192	2026-10-01	community	–	source

All rows cite their source. Provenance: community 40018 sourced 1085