Llama-3.2-3B-Instruct

16 records across 3 hardware.

hardware	model	quant	backend	tok/s	pp tok/s	W	ttft s	ctx/tg	date	class	flags	source
M4 Max	Llama-3.2-3B-Instruct	4bit	oMLX	135.4	1615.0	–	–	4096	2026-10-10	community	–	source
M5 Pro	Llama-3.2-3B-Instruct	4bit	oMLX	109.6	4203.0	–	–	4096	2026-10-11	community	–	source
M5 Pro	Llama-3.2-3B-Instruct	4bit	oMLX	109.6	4203.0	–	–	4096	2026-10-11	community	–	source
M4 Max	Llama-3.2-3B-Instruct	4bit	oMLX	88.7	1188.0	–	–	16384	2026-10-10	community	–	source
M4 Max	Llama-3.2-3B-Instruct	8bit	oMLX	84.4	1763.0	–	–	4096	2026-10-09	community	–	source
M4 Max	Llama-3.2-3B-Instruct	8bit	oMLX	76.7	1585.0	–	–	8192	2026-10-09	community	–	source
M5 Pro	Llama-3.2-3B-Instruct	4bit	oMLX	68.7	3278.0	–	–	16384	2026-10-11	community	–	source
M5 Pro	Llama-3.2-3B-Instruct	4bit	oMLX	68.7	3278.0	–	–	16384	2026-10-11	community	–	source
M4 Max	Llama-3.2-3B-Instruct	8bit	oMLX	68.2	1669.0	–	–	4096	2026-10-09	community	contradiction	source
M4 Max	Llama-3.2-3B-Instruct	8bit	oMLX	68.2	1562.0	–	–	8192	2026-10-09	community	contradiction	source
M4 Max	Llama-3.2-3B-Instruct	8bit	oMLX	64.1	1309.0	–	–	16384	2026-10-09	community	–	source
M4 Max	Llama-3.2-3B-Instruct	8bit	oMLX	64.1	1315.0	–	–	16384	2026-10-09	community	–	source
M4	Llama-3.2-3B-Instruct	4bit	oMLX	45.2	520.4	–	–	1024	2026-03-09	community	–	source
M4	Llama-3.2-3B-Instruct	4bit	oMLX	45.2	520.4	–	–	1024	2026-03-09	community	–	source
M4	Llama-3.2-3B-Instruct	4bit	oMLX	37.0	478.6	–	–	4096	2026-03-09	community	–	source
M4	Llama-3.2-3B-Instruct	4bit	oMLX	37.0	478.6	–	–	4096	2026-03-09	community	–	source