Qwen3.6-27B

131 records across 20 hardware.

hardware	model	quant	backend	tok/s	pp tok/s	W	ttft s	ctx/tg	date	class	flags	source
RTX PRO 6000 Blackwell Workstation Edition	Qwen3.6-27B	Q4_K_M	llama.cpp	145.14	–	–	–	–	2026-10-12	sourced	–	source
RTX 5090	Qwen3.6-27B	Q4_K_M	llama.cpp	136.59	–	–	–	–	2026-10-12	sourced	–	source
RTX PRO 6000 Blackwell Workstation Edition	Qwen3.6-27B	Q4_K_M	llama.cpp	135.98	–	–	–	–	2026-10-12	sourced	–	source
RTX 5090	Qwen3.6-27B	Q4_K_M	llama.cpp	130.11	–	–	–	–	2026-10-12	sourced	–	source
RTX 5090	Qwen3.6-27B 27B hybrid GDN + full attention (MTP head)	NVFP4	vLLM	95.6	–	–	–	–	2026-07	community	–	source
5090	Qwen3.6-27B	Q4_K_S MTP	LM Studio	90	–	–	–	–	2026-07-17	community	–	source
5090	Qwen3.6-27B	Q4_K_S	LM Studio	75	–	–	–	–	2026-07-17	community	–	source
M5 Max	Qwen3.6-27B	4bit	oMLX	66.4	993.7	–	–	4096	2026-10-09	community	–	source
M5 Max	Qwen3.6-27B	4bit	oMLX	66.2	902.4	–	–	1024	2026-10-09	community	–	source
M5 Max	Qwen3.6-27B	4bit	oMLX	58.3	831.1	–	–	1024	2026-10-09	community	contradiction	source
M5 Max	Qwen3.6-27B	4bit	oMLX	57.4	897.0	–	–	1024	2026-10-09	community	contradiction	source
M5 Max	Qwen3.6-27B	oQ6e	oMLX	56.3	709.9	–	–	32768	2026-10-11	community	–	source
M5 Max	Qwen3.6-27B	4bit	oMLX	55.9	942.4	–	–	4096	2026-10-09	community	contradiction	source
M5 Max	Qwen3.6-27B	4bit	oMLX	52.9	990.1	–	–	4096	2026-10-09	community	contradiction	source
M5 Max	Qwen3.6-27B	oQ6e	oMLX	44.8	928.5	–	–	4096	2026-10-11	community	–	source
M3 Max	Qwen3.6-27B	4bit	oMLX	36.8	166.9	–	–	1024	2026-10-09	community	–	source
M5 Max	Qwen3.6-27B	oQ6e	oMLX	34.7	561.2	–	–	65536	2026-10-11	community	–	source
M5 Max	Qwen3.6-27B	4bit	oMLX	33.0	992.3	–	–	4096	2026-10-11	community	contradiction	source
M3 Ultra	Qwen3.6-27B	OptiQ-4bit	oMLX	32.8	345.8	–	–	4096	2026-10-07	community	–	source
M3 Ultra	Qwen3.6-27B	OptiQ-4bit	oMLX	32.8	348.8	–	–	4096	2026-10-08	community	–	source
M3 Max	Qwen3.6-27B	4bit	oMLX	32.8	165.8	–	–	1024	2026-10-09	community	contradiction	source
M2 Ultra	Qwen3.6-27B	4bit	oMLX	32.1	202.4	–	–	4096	2026-10-07	community	–	source
M5 Max	Qwen3.6-27B	4bit	oMLX	31.0	892.1	–	–	16384	2026-10-11	community	–	source
M2 Ultra	Qwen3.6-27B	4bit	oMLX	29.6	187.7	–	–	16384	2026-10-07	community	–	source
M4 Max	Qwen3.6-27B	4bit	oMLX	29.2	261.3	–	–	1024	2026-10-06	community	–	source
M4 Max	Qwen3.6-27B	4bit	oMLX	29.2	261.3	–	–	1024	2026-10-06	community	–	source
M5 Max	Qwen3.6-27B	OptiQ-4bit	oMLX	28.2	915.7	–	–	1024	2026-10-10	community	–	source
M5 Max	Qwen3.6-27B	OptiQ-4bit	oMLX	28.0	1009.0	–	–	4096	2026-10-10	community	–	source
M3 Max	Qwen3.6-27B	4bit	oMLX	27.4	174.2	–	–	4096	2026-10-09	community	–	source
M4 Max	Qwen3.6-27B	4bit	oMLX	26.8	232.9	–	–	4096	2026-10-06	community	–	source
M4 Max	Qwen3.6-27B	4bit	oMLX	26.8	232.9	–	–	4096	2026-10-06	community	–	source
GB10	Qwen3.6-27B	Q4_K_M	llama.cpp	26.77	–	–	–	–	2026-10-12	sourced	–	source
M5 Max	Qwen3.6-27B	oQ4e	oMLX	25.9	1321.0	–	–	8192	2026-10-17	community	–	source
M3 Max	Qwen3.6-27B	4bit	oMLX	25.4	173.8	–	–	4096	2026-10-09	community	–	source
GB10	Qwen3.6-27B	Q4_K_M	llama.cpp	24.29	–	–	–	–	2026-10-12	sourced	–	source
M5 Max	Qwen3.6-27B	bf16	oMLX	21.7	790.1	–	–	32768	2026-10-12	community	–	source
M5 Max	Qwen3.6-27B	bf16	oMLX	20.0	923.0	–	–	4096	2026-10-12	community	–	source
M5 Max	Qwen3.6-27B	bf16	oMLX	19.9	988.2	–	–	8192	2026-10-12	community	–	source
M2 Max	Qwen3.6-27B	4bit	oMLX	18.9	97.7	–	–	4096	2026-10-06	community	–	source
M2 Max	Qwen3.6-27B	4bit	oMLX	18.9	97.7	–	–	4096	2026-10-06	community	–	source
M5 Max	Qwen3.6-27B	bf16	oMLX	18.6	825.4	–	–	16384	2026-10-12	community	–	source
M5 Pro	Qwen3.6-27B	4bit	oMLX	18.0	475.1	–	–	1024	2026-10-08	community	–	source
M5 Max	Qwen3.6-27B	bf16	oMLX	17.9	616.1	–	–	1024	2026-10-12	community	–	source
M5 Pro	Qwen3.6-27B	4bit	oMLX	17.7	487.0	–	–	4096	2026-10-08	community	–	source
M2 Max	Qwen3.6-27B	4bit	oMLX	17.6	93.8	–	–	16384	2026-10-06	community	–	source
M2 Max	Qwen3.6-27B	4bit	oMLX	17.6	93.8	–	–	16384	2026-10-06	community	–	source
M4 Max (128 GB)	Qwen3.6-27B	Q4_K - Medium	Ollama	16.56	114.49	–	–	–	–	sourced	–	source
M4 Max	Qwen3.6-27B	8bit	oMLX	16.3	256.6	–	–	4096	2026-10-09	community	–	source
M4 Max	Qwen3.6-27B	8bit	oMLX	16.2	255.8	–	–	1024	2026-10-06	community	–	source
M4 Max	Qwen3.6-27B	8bit	oMLX	16.2	255.8	–	–	1024	2026-10-06	community	–	source
M4 Max	Qwen3.6-27B	8bit	oMLX	15.7	244.9	–	–	16384	2026-10-09	community	–	source
M5 Max	Qwen3.6-27B	8bit	oMLX	15.6	852.9	–	–	4096	2026-10-14	community	–	source
M5 Max	Qwen3.6-27B	8bit	oMLX	15.6	852.9	–	–	4096	2026-10-14	community	–	source
M2 Max	Qwen3.6-27B	4bit	oMLX	15.6	135.6	–	–	1024	2026-10-15	community	–	source
M2 Max	Qwen3.6-27B	4bit	oMLX	15.6	136.7	–	–	1024	2026-10-15	community	–	source
M4 Max	Qwen3.6-27B	8bit	oMLX	15.5	213.1	–	–	4096	2026-10-06	community	–	source
M4 Max	Qwen3.6-27B	8bit	oMLX	15.5	213.1	–	–	4096	2026-10-06	community	–	source
M5 Max	Qwen3.6-27B	bf16	oMLX	15.2	700.0	–	–	65536	2026-10-12	community	–	source
M5 Max	Qwen3.6-27B	8bit	oMLX	15.2	888.3	–	–	16384	2026-10-14	community	–	source
M5 Max	Qwen3.6-27B	8bit	oMLX	15.2	888.3	–	–	16384	2026-10-14	community	–	source
M2 Max	Qwen3.6-27B	4bit	oMLX	14.9	124.1	–	–	4096	2026-10-15	community	–	source
M1 Pro	Qwen3.6-27B	oQ4	oMLX	14.9	78.2	–	–	4096	2026-10-16	community	–	source
M1 Pro	Qwen3.6-27B	oQ4	oMLX	14.7	76.2	–	–	1024	2026-10-16	community	–	source
M1 Pro	Qwen3.6-27B	oQ4	oMLX	14.6	74.0	–	–	16384	2026-10-16	community	–	source
M1 Pro	Qwen3.6-27B	oQ4	oMLX	13.9	76.0	–	–	1024	2026-10-16	community	–	source
M4 Max	Qwen3.6-27B	8bit	oMLX	13.9	200.2	–	–	65536	2026-10-18	community	–	source
M5 Pro	Qwen3.6-27B	Q4_K_XL	oMLX	13.4	467.2	–	–	1024	2026-10-07	community	–	source
M1 Pro	Qwen3.6-27B	oQ4	oMLX	13.4	78.6	–	–	4096	2026-10-16	community	contradiction	source
M5 Pro	Qwen3.6-27B	6bit	oMLX	13.0	440.8	–	–	1024	2026-10-07	community	–	source
M3 Max	Qwen3.6-27B	8bit	oMLX	12.9	219.9	–	–	1024	2026-10-07	community	–	source
M4 Pro	Qwen3.6-27B	4bit	oMLX	12.9	131.3	–	–	4096	2026-10-12	community	–	source
M3 Max	Qwen3.6-27B	8bit	oMLX	12.8	225.3	–	–	4096	2026-10-07	community	–	source
M4 Pro	Qwen3.6-27B	4bit	oMLX	12.8	132.3	–	–	4096	2026-10-12	community	–	source
M5 Pro	Qwen3.6-27B	6bit	oMLX	12.7	484.3	–	–	4096	2026-10-07	community	–	source
M5 Pro	Qwen3.6-27B	Q4_K_XL	oMLX	12.7	375.3	–	–	1024	2026-10-07	community	–	source
M5 Pro	Qwen3.6-27B	Q4_K_XL	oMLX	12.7	469.8	–	–	4096	2026-10-07	community	–	source
M5 Pro	Qwen3.6-27B	Q4_K_XL	oMLX	12.7	505.6	–	–	4096	2026-10-07	community	–	source
M5 Max	Qwen3.6-27B	bf16	oMLX	12.3	576.9	–	–	131072	2026-10-12	community	–	source
M1 Max	Qwen3.6-27B	OptiQ-4bit	oMLX	12.1	81.4	–	–	4096	2026-10-07	community	–	source
M4 Pro	Qwen3.6-27B	4bit	oMLX	12.1	126.8	–	–	16384	2026-10-12	community	–	source
M4 Pro	Qwen3.6-27B	4bit	oMLX	12.1	127.1	–	–	16384	2026-10-12	community	–	source
M4 Max	Qwen3.6-27B	8bit	oMLX	11.5	142.0	–	–	131072	2026-10-18	community	–	source
M2 Pro	Qwen3.6-27B	oQ4	oMLX	11.4	85.6	–	–	1024	2026-10-09	community	–	source
M1 Max	Qwen3.6-27B	OptiQ-4bit	oMLX	11.2	77.4	–	–	16384	2026-10-07	community	–	source
gfx1151	Qwen3.6-27B	AWQ-INT4	vLLM	11.0	–	–	–	128	2026-10-13	sourced	–	source
M2 Pro	Qwen3.6-27B	oQ4	oMLX	11.0	88.5	–	–	4096	2026-10-09	community	–	source
gfx1151	Qwen3.6-27B	AWQ-INT4	vLLM	10.9	–	–	–	128	2026-10-13	sourced	–	source
M5 Pro	Qwen3.6-27B	6bit	oMLX	10.9	430.1	–	–	32768	2026-10-07	community	–	source
gfx1151	Qwen3.6-27B	AWQ-INT4	vLLM	10.8	–	–	–	128	2026-10-13	sourced	–	source
gfx1151	Qwen3.6-27B	AWQ-INT4	vLLM	10.7	–	–	–	128	2026-10-13	sourced	–	source
gfx1151	Qwen3.6-27B	AWQ-INT4	vLLM	10.7	–	–	–	128	2026-10-13	sourced	–	source
gfx1151	Qwen3.6-27B	AWQ-INT4	vLLM	10.7	–	–	–	128	2026-10-13	sourced	–	source
M5 Pro	Qwen3.6-27B	4bit	oMLX	10.7	189.7	–	–	1024	2026-10-18	community	contradiction	source
M5 Pro	Qwen3.6-27B	4bit	oMLX	10.7	240.3	–	–	4096	2026-10-18	community	contradiction	source
M5 Pro	Qwen3.6-27B	4bit	oMLX	10.7	189.7	–	–	1024	2026-10-18	community	contradiction	source
M5 Pro	Qwen3.6-27B	4bit	oMLX	10.7	240.3	–	–	4096	2026-10-18	community	contradiction	source
gfx1151	Qwen3.6-27B	AWQ-INT4	vLLM	10.6	–	–	–	8192	2026-10-13	sourced	–	source
gfx1151	Qwen3.6-27B	AWQ-INT4	vLLM	10.5	–	–	–	8192	2026-10-13	sourced	–	source
gfx1151	Qwen3.6-27B	AWQ-INT4	vLLM	10.4	–	–	–	128	2026-10-13	sourced	–	source
gfx1151	Qwen3.6-27B	AWQ-INT4	vLLM	10.4	–	–	–	8192	2026-10-13	sourced	–	source
gfx1151	Qwen3.6-27B	AWQ-INT4	vLLM	10.3	–	–	–	8192	2026-10-13	sourced	–	source
M5 Max	Qwen3.6-27B	bf16	oMLX	10.3	479.6	–	–	199680	2026-10-12	community	–	source
M1 Pro	Qwen3.6-27B	oQ4e	oMLX	9.9	44.6	–	–	1024	2026-10-07	community	–	source
M1 Pro	Qwen3.6-27B	oQ4e	oMLX	9.7	47.3	–	–	4096	2026-10-07	community	–	source
M6	Qwen3.6-27B	4bit	oMLX	9.7	291.2	–	–	1024	2026-10-11	community	–	source
M6	Qwen3.6-27B	4bit	oMLX	9.7	291.2	–	–	1024	2026-10-11	community	–	source
M5 Pro	Qwen3.6-27B	6bit	oMLX	9.6	372.0	–	–	65536	2026-10-07	community	–	source
M6	Qwen3.6-27B	4bit	oMLX	9.3	296.1	–	–	4096	2026-10-11	community	–	source
M6	Qwen3.6-27B	4bit	oMLX	9.3	296.1	–	–	4096	2026-10-11	community	–	source
M3 Pro	Qwen3.6-27B	4bit	oMLX	8.8	102.3	–	–	1024	2026-10-14	community	–	source
M4 Pro	Qwen3.6-27B	8bit	oMLX	8.7	104.5	–	–	1024	2026-10-07	community	–	source
M1 Pro	Qwen3.6-27B	oQ4	oMLX	8.7	76.4	–	–	16384	2026-10-16	community	contradiction	source
M4 Pro	Qwen3.6-27B	8bit	oMLX	8.6	106.0	–	–	4096	2026-10-07	community	–	source
M3 Pro	Qwen3.6-27B	4bit	oMLX	8.6	103.5	–	–	4096	2026-10-18	community	–	source
gfx1151	Qwen3.6-27B	AWQ-INT4	vLLM	8.5	–	–	–	128	2026-10-13	sourced	contradiction	source
M4 Pro	Qwen3.6-27B	8bit	oMLX	8.3	96.5	–	–	16384	2026-10-07	community	–	source
M3 Pro	Qwen3.6-27B	4bit	oMLX	8.3	102.0	–	–	1024	2026-10-18	community	–	source
RTX 3060 + Intel Arc A770	Qwen3.6-27B 27B	Q4_K_M	llama.cpp (Vulkan)	7.88	97.37	–	–	–	2026-08	community	–	source
M4 Pro	Qwen3.6-27B	8bit	oMLX	7.8	93.6	–	–	32768	2026-10-07	community	–	source
M3 Pro	Qwen3.6-27B	4bit	oMLX	6.5	89.9	–	–	1024	2026-10-13	community	contradiction	source
M5 Pro	Qwen3.6-27B	8bit	oMLX	6.1	172.2	–	–	1024	2026-10-18	community	–	source
M5 Pro	Qwen3.6-27B	8bit	oMLX	6.1	212.0	–	–	4096	2026-10-18	community	–	source
M5 Pro	Qwen3.6-27B	8bit	oMLX	6.1	172.2	–	–	1024	2026-10-18	community	–	source
M5 Pro	Qwen3.6-27B	8bit	oMLX	6.1	212.0	–	–	4096	2026-10-18	community	–	source
gfx1151	Qwen3.6-27B	AWQ-INT4	vLLM	5.5	–	–	–	128	2026-10-13	sourced	contradiction	source
gfx1151	Qwen3.6-27B	AWQ-INT4	vLLM	4.8	–	–	–	8192	2026-10-13	sourced	–	source
gfx1151	Qwen3.6-27B	AWQ-INT4	vLLM	4.0	–	–	–	8192	2026-10-13	sourced	contradiction	source
gfx1151	Qwen3.6-27B	AWQ-INT4	vLLM	4.0	–	–	–	8192	2026-10-13	sourced	contradiction	source
gfx1151	Qwen3.6-27B	AWQ-INT4	vLLM	3.7	–	–	–	8192	2026-10-13	sourced	contradiction	source
RTX 3060 + Intel Arc A770	Qwen3.6-27B 27B	Q4_K_M	llama.cpp (Vulkan)	2.93	98.26	–	–	–	2026-08	community	–	source
gfx1151	Qwen3.6-27B	AWQ-INT4	vLLM	2.8	–	–	–	8192	2026-10-13	sourced	contradiction	source

← all models