gpt-oss 20B

362 records across 33 hardware.

hardware	model	quant	backend	tok/s	pp tok/s	W	ttft s	ctx/tg	date	class	flags	source
sm_80	gpt-oss 20B	unspecified	vLLM	11546	–	–	–	1024	2026-10-03	community	–	source
sm_80	gpt-oss 20B	unspecified	vLLM	11283	–	–	–	1024	2026-10-03	community	–	source
A100	gpt-oss 20B 20B MoE	as-published	vLLM	914.97	–	–	–	–	2026-09	community	–	source
A100	gpt-oss 20B 20B MoE	as-published	vLLM	501.49	–	–	–	–	2026-09	community	–	source
AMD Zen 5 (Turin, 32 cores pinned to one NUMA node)	gpt-oss 20B	BF16-w8a8	vllm	375.74	–	–	–	128	2026-09-30	community	–	source
AMD Zen 5 (Turin, 32 cores pinned to one NUMA node)	gpt-oss 20B	BF16-w8a8	vllm	369.79	–	–	–	128	2026-09-30	community	–	source
RTX 5090	gpt-oss 20B 20B	Q4_K	llama.cpp	298.2	9443.8	–	–	4096	–	sourced	–	source
RTX 5090	gpt-oss 20B 20B	Q4_K	llama.cpp	249.2	7168.1	–	–	16384	–	sourced	–	source
RTX PRO 4500 Blackwell	gpt-oss 20B	Q4_K - Medium	llama.cpp (CUDA)	240.75	6760.29	–	–	tg128	2025-11-14	community	–	source
RTX PRO 4500 Blackwell	gpt-oss 20B	Q4_K - Medium	llama.cpp (CUDA)	233.97	5812.88	–	–	tg128	2025-11-14	community	–	source
RTX 5090	gpt-oss 20B 20B	Q4_K	llama.cpp	215.1	5183.1	–	–	32768	–	sourced	–	source
RTX Pro 5000 Blackwell	gpt-oss 20B 20B	MXFP4	llama.cpp	213.4	10339.6	–	–	4096	–	sourced	–	source
RTX Pro 5000 Blackwell	gpt-oss 20B 20B	MXFP4	llama.cpp	196.2	8747.7	–	–	16384	–	sourced	–	source
RTX 4090	gpt-oss 20B 20B	MXFP4	llama.cpp	190.6	8369.3	–	–	4096	–	sourced	–	source
RTX Pro 5000 Blackwell	gpt-oss 20B 20B	MXFP4	llama.cpp	182.8	7257.7	–	–	32768	–	sourced	–	source
RTX 5080	gpt-oss 20B 20B	MXFP4	llama.cpp	172.4	–	–	–	4096	–	sourced	–	source
RTX 5090	gpt-oss 20B 20B	Q4_K	llama.cpp	169.0	3019.7	–	–	65536	–	sourced	–	source
RTX 4090	gpt-oss 20B 20B	MXFP4	llama.cpp	163.3	5856.2	–	–	16384	–	sourced	–	source
RTX Pro 5000 Blackwell	gpt-oss 20B 20B	MXFP4	llama.cpp	161.4	5273.9	–	–	65536	–	sourced	–	source
RTX 5080	gpt-oss 20B 20B	MXFP4	llama.cpp	161.3	–	–	–	16384	–	sourced	–	source
RTX 6000 Ada	gpt-oss 20B 20B	MXFP4	llama.cpp	160.8	6495.8	–	–	4096	–	sourced	–	source
RTX 3090 Ti	gpt-oss 20B 20B	MXFP4	llama.cpp	160.3	4805.4	–	–	4096	–	sourced	–	source
AMD Zen 5 (Turin, 32 cores pinned to one NUMA node)	gpt-oss 20B	BF16-w8a8	vllm	158.21	–	–	–	128	2026-09-30	community	–	source
RTX 5070 Ti	gpt-oss 20B 20B	MXFP4	llama.cpp	156.0	6178.3	–	–	4096	–	sourced	–	source
AMD Zen 5 (Turin, 32 cores pinned to one NUMA node)	gpt-oss 20B	BF16-w8a8	vllm	154.23	–	–	–	128	2026-09-30	community	–	source
RTX 5080	gpt-oss 20B 20B	MXFP4	llama.cpp	149.3	–	–	–	32768	–	sourced	–	source
RX 6800	gpt-oss 20B 23.6B	MXFP4	llama.cpp (Vulkan)	145.9	1883	–	–	tg128	2026-10-01	community	–	source
RTX 4090	gpt-oss 20B 20B	MXFP4	llama.cpp	140.7	3705.6	–	–	32768	–	sourced	–	source
RTX 4080 SUPER	gpt-oss 20B 20B	MXFP4	llama.cpp	139.1	6364.0	–	–	4096	–	sourced	–	source
RTX 3090 Ti	gpt-oss 20B 20B	MXFP4	llama.cpp	137.6	3440.1	–	–	16384	–	sourced	–	source
M5 Max	gpt-oss 20B	MXFP4-Q8	oMLX	137.6	1268.0	–	–	1024	2026-10-11	community	–	source
RTX 6000 Ada	gpt-oss 20B 20B	MXFP4	llama.cpp	137.1	5350.5	–	–	16384	–	sourced	–	source
M5 Max	gpt-oss 20B	MXFP4-Q8	oMLX	137.1	4964.0	–	–	1024	2026-10-11	community	–	source
RTX 4080	gpt-oss 20B 20B	MXFP4	llama.cpp	136.5	6218.5	–	–	4096	–	sourced	–	source
RTX 5070 Ti	gpt-oss 20B 20B	MXFP4	llama.cpp	133.1	4940.7	–	–	16384	–	sourced	–	source
M5 Max	gpt-oss 20B	MXFP4-Q8	oMLX	132.6	1255.0	–	–	1024	2026-10-18	community	–	source
RTX 5080	gpt-oss 20B 20B	MXFP4	llama.cpp	130.9	–	–	–	65536	–	sourced	–	source
RTX Pro 5000 Blackwell	gpt-oss 20B 20B	MXFP4	llama.cpp	130.9	3163.3	–	–	131072	–	sourced	–	source
M3 Ultra	gpt-oss 20B	MXFP4-Q8	oMLX	130.2	1577.0	–	–	1024	2026-10-10	community	–	source
M5 Max	gpt-oss 20B	MXFP4-Q8	oMLX	130.0	4712.0	–	–	4096	2026-10-11	community	–	source
RTX 4070 Ti SUPER	gpt-oss 20B 20B	MXFP4	llama.cpp	128.9	5628.1	–	–	4096	–	sourced	–	source
M3 Ultra	gpt-oss 20B	MXFP4-Q8	oMLX	124.9	2472.0	–	–	4096	2026-10-10	community	–	source
M5 Max	gpt-oss 20B	MXFP4-Q8	oMLX	123.9	4520.0	–	–	4096	2026-10-18	community	–	source
RTX 4080 SUPER	gpt-oss 20B 20B	MXFP4	llama.cpp	123.0	4708.7	–	–	16384	–	sourced	–	source
M4 Max	gpt-oss 20B	MXFP4-Q8	oMLX	122.3	1226.0	–	–	1024	2026-10-15	community	–	source
RX 6800	gpt-oss 20B 23.6B	MXFP4	llama.cpp (ROCm)	121.2	2022	–	–	tg128	2026-10-01	community	–	source
RTX 4080	gpt-oss 20B 20B	MXFP4	llama.cpp	120.2	4329.2	–	–	16384	–	sourced	–	source
RTX 3090 Ti	gpt-oss 20B 20B	MXFP4	llama.cpp	119.7	2521.3	–	–	32768	–	sourced	–	source
M4 Max	gpt-oss 20B	MXFP4-Q8	oMLX	118.7	1652.0	–	–	4096	2026-10-15	community	–	source
RTX 6000 Ada	gpt-oss 20B 20B	MXFP4	llama.cpp	116.3	3818.7	–	–	32768	–	sourced	–	source
RTX 5070 Ti	gpt-oss 20B 20B	MXFP4	llama.cpp	115.9	3903.5	–	–	32768	–	sourced	–	source
RTX 4070 Ti SUPER	gpt-oss 20B 20B	MXFP4	llama.cpp	113.6	4182.7	–	–	16384	–	sourced	–	source
RTX 5090	gpt-oss 20B 20B	Q4_K	llama.cpp	112.0	1636.4	–	–	131072	–	sourced	–	source
RTX A6000	gpt-oss 20B 20B	MXFP4	llama.cpp	108.4	3490.6	–	–	4096	–	sourced	–	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	108.1	1368.0	–	–	1024	2026-10-06	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	108.1	1368.0	–	–	1024	2026-10-06	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	107.3	1361.0	–	–	1024	2026-10-06	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	107.3	1361.0	–	–	1024	2026-10-06	community	–	source
RTX 4080 SUPER	gpt-oss 20B 20B	MXFP4	llama.cpp	107.2	3328.1	–	–	32768	–	sourced	–	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	107.0	1368.0	–	–	1024	2026-10-06	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	107.0	1368.0	–	–	1024	2026-10-06	community	–	source
Arc Pro B70	gpt-oss 20B	MXFP4	llama.cpp (SYCL)	106.64	–	–	–	–	2026-10-01	community	–	source
RTX 4080	gpt-oss 20B 20B	MXFP4	llama.cpp	106.0	2831.1	–	–	32768	–	sourced	–	source
Radeon RX 9060 XT	gpt-oss 20B	MXFP4	llama.cpp (Vulkan)	105.8	3464	–	–	tg128	2026-10-02	community	–	source
RTX 5080	gpt-oss 20B 20B	MXFP4	llama.cpp	105.6	–	–	–	131072	–	sourced	–	source
M5 Max	gpt-oss 20B	MXFP4-Q8	oMLX	105.1	4274.0	–	–	4096	2026-10-12	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	105.0	1017.0	–	–	1024	2026-10-06	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	105.0	1017.0	–	–	1024	2026-10-06	community	–	source
M5 Max	gpt-oss 20B	MXFP4-Q8	oMLX	104.9	4287.0	–	–	4096	2026-10-18	community	–	source
M5 Max	gpt-oss 20B	MXFP4-Q8	oMLX	104.6	4204.0	–	–	4096	2026-10-18	community	–	source
M5 Max	gpt-oss 20B	MXFP4-Q8	oMLX	104.5	4223.0	–	–	4096	2026-10-18	community	–	source
RTX 4090	gpt-oss 20B 20B	MXFP4	llama.cpp	104.0	2092.2	–	–	65536	–	sourced	–	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	103.5	1348.0	–	–	4096	2026-10-06	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	103.5	1348.0	–	–	4096	2026-10-06	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	103.3	1351.0	–	–	4096	2026-10-06	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	103.3	1351.0	–	–	4096	2026-10-06	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	102.7	1374.0	–	–	4096	2026-10-06	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	102.7	1374.0	–	–	4096	2026-10-06	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	100.3	1369.0	–	–	1024	2026-10-06	community	–	source
M5 Pro	gpt-oss 20B	MXFP4-Q4	oMLX	99.8	2300.0	–	–	1024	2026-10-12	community	–	source
M5 Pro	gpt-oss 20B	MXFP4-Q4	oMLX	99.8	2300.0	–	–	1024	2026-10-12	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	99.3	1369.0	–	–	1024	2026-10-06	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	99.3	1369.0	–	–	1024	2026-10-06	community	–	source
RTX 4070 Ti SUPER	gpt-oss 20B 20B	MXFP4	llama.cpp	98.7	3020.0	–	–	32768	–	sourced	–	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	98.7	1372.0	–	–	1024	2026-10-06	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	98.7	1372.0	–	–	1024	2026-10-06	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	96.8	1376.0	–	–	4096	2026-10-06	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	96.8	1376.0	–	–	4096	2026-10-06	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	96.1	1378.0	–	–	4096	2026-10-06	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	96.1	1378.0	–	–	4096	2026-10-06	community	–	source
RTX 3090 Ti	gpt-oss 20B 20B	MXFP4	llama.cpp	94.9	1666.5	–	–	65536	–	sourced	–	source
M5 Pro	gpt-oss 20B	OptiQ-4bit	oMLX	94.8	2430.0	–	–	1024	2026-10-14	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	94.7	1155.0	–	–	4096	2026-10-06	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	94.7	1155.0	–	–	4096	2026-10-06	community	–	source
M5 Pro	gpt-oss 20B	OptiQ-4bit	oMLX	94.6	2427.0	–	–	1024	2026-10-14	community	–	source
M5 Pro	gpt-oss 20B	OptiQ-4bit	oMLX	94.6	2427.0	–	–	1024	2026-10-14	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	94.1	839.9	–	–	1024	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	94.1	839.9	–	–	1024	2026-10-06	community	contradiction	source
M5 Pro	gpt-oss 20B	MXFP4-Q4	oMLX	94.0	2449.0	–	–	4096	2026-10-12	community	–	source
M5 Pro	gpt-oss 20B	MXFP4-Q4	oMLX	94.0	2449.0	–	–	4096	2026-10-12	community	–	source
Radeon RX 9060 XT	gpt-oss 20B	MXFP4	llama.cpp (ROCm)	92.8	3372	–	–	tg128	2026-10-02	community	–	source
RTX 5060 Ti 16GB	gpt-oss 20B 20B	MXFP4	llama.cpp	92.1	3585.2	–	–	4096	–	sourced	–	source
RTX 5070 Ti	gpt-oss 20B 20B	MXFP4	llama.cpp	91.3	2644.9	–	–	65536	–	sourced	–	source
RTX A6000	gpt-oss 20B 20B	MXFP4	llama.cpp	91.2	2649.2	–	–	16384	–	sourced	–	source
M5 Max	gpt-oss 20B	MXFP4-Q8	oMLX	90.5	4029.0	–	–	16384	2026-10-18	community	–	source
M5 Max	gpt-oss 20B	MXFP4-Q8	oMLX	90.5	4048.0	–	–	16384	2026-10-18	community	–	source
M5 Max	gpt-oss 20B	MXFP4-Q8	oMLX	90.2	4019.0	–	–	16384	2026-10-18	community	–	source
M5 Max	gpt-oss 20B	MXFP4-Q8	oMLX	90.1	4018.0	–	–	16384	2026-10-12	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	89.7	975.0	–	–	1024	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	89.7	975.0	–	–	1024	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	89.6	963.4	–	–	1024	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	89.6	963.4	–	–	1024	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	89.6	1058.0	–	–	1024	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	89.6	1058.0	–	–	1024	2026-10-06	community	contradiction	source
M5 Pro	gpt-oss 20B	OptiQ-4bit	oMLX	89.1	2518.0	–	–	4096	2026-10-14	community	–	source
M5 Pro	gpt-oss 20B	OptiQ-4bit	oMLX	89.1	2518.0	–	–	4096	2026-10-14	community	–	source
M5 Pro	gpt-oss 20B	OptiQ-4bit	oMLX	89.0	2526.0	–	–	4096	2026-10-14	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	88.7	1072.0	–	–	4096	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	88.7	1072.0	–	–	4096	2026-10-06	community	contradiction	source
RTX 6000 Ada	gpt-oss 20B 20B	MXFP4	llama.cpp	88.4	2270.3	–	–	65536	–	sourced	–	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	87.0	906.8	–	–	1024	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	87.0	906.8	–	–	1024	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	86.4	996.6	–	–	1024	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	86.4	996.6	–	–	1024	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	85.9	1072.0	–	–	4096	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	85.9	1072.0	–	–	4096	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	85.8	1039.0	–	–	1024	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	85.8	1039.0	–	–	1024	2026-10-06	community	contradiction	source
M5 Pro	gpt-oss 20B	MXFP4-Q4	oMLX	85.4	2432.0	–	–	8192	2026-10-12	community	–	source
M5 Pro	gpt-oss 20B	MXFP4-Q4	oMLX	85.4	2432.0	–	–	8192	2026-10-12	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	84.9	989.2	–	–	4096	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	84.9	989.2	–	–	4096	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	84.7	965.9	–	–	4096	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	84.7	965.9	–	–	4096	2026-10-06	community	contradiction	source
M5 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	84.3	2423.0	–	–	1024	2026-10-15	community	–	source
M5 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	84.2	983.5	–	–	1024	2026-10-15	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	84.0	897.2	–	–	1024	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	84.0	897.2	–	–	1024	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	83.9	918.0	–	–	4096	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	83.9	918.0	–	–	4096	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	83.8	905.6	–	–	1024	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	83.8	905.6	–	–	1024	2026-10-06	community	contradiction	source
M5 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	83.4	853.1	–	–	1024	2026-10-11	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	83.1	1025.0	–	–	4096	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	83.1	1025.0	–	–	4096	2026-10-06	community	contradiction	source
RTX 5060 Ti 16GB	gpt-oss 20B 20B	MXFP4	llama.cpp	82.4	2753.3	–	–	16384	–	sourced	–	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	82.4	1086.0	–	–	4096	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	82.4	1086.0	–	–	4096	2026-10-06	community	contradiction	source
M5 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	81.6	2397.0	–	–	1024	2026-10-10	community	–	source
RTX 4080 SUPER	gpt-oss 20B 20B	MXFP4	llama.cpp	81.5	1961.5	–	–	65536	–	sourced	–	source
M5 Pro	gpt-oss 20B	OptiQ-4bit	oMLX	81.5	2440.0	–	–	8192	2026-10-14	community	–	source
M5 Pro	gpt-oss 20B	OptiQ-4bit	oMLX	81.5	2431.0	–	–	8192	2026-10-14	community	–	source
M5 Pro	gpt-oss 20B	OptiQ-4bit	oMLX	81.5	2431.0	–	–	8192	2026-10-14	community	–	source
M5 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	81.4	2831.0	–	–	4096	2026-10-18	community	–	source
M5 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	81.4	2831.0	–	–	4096	2026-10-18	community	–	source
RTX 4080	gpt-oss 20B 20B	MXFP4	llama.cpp	81.3	1559.0	–	–	65536	–	sourced	–	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	81.3	1013.0	–	–	1024	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	81.3	1013.0	–	–	1024	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	80.5	1020.0	–	–	4096	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	80.5	1020.0	–	–	4096	2026-10-06	community	contradiction	source
M5 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	80.3	2354.0	–	–	4096	2026-10-15	community	–	source
M5 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	79.9	1899.0	–	–	4096	2026-10-09	community	–	source
M5 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	79.7	2530.0	–	–	4096	2026-10-15	community	–	source
M5 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	79.4	2377.0	–	–	1024	2026-10-10	community	–	source
RTX 4070 Ti SUPER	gpt-oss 20B 20B	MXFP4	llama.cpp	78.7	1716.2	–	–	65536	–	sourced	–	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	78.3	1047.0	–	–	4096	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	78.3	1047.0	–	–	4096	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	77.1	878.5	–	–	1024	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	77.1	878.5	–	–	1024	2026-10-06	community	contradiction	source
M2 Max	gpt-oss 20B	MXFP4-Q8	oMLX	76.1	541.3	–	–	1024	2026-10-08	community	–	source
M5 Max	gpt-oss 20B	MXFP4-Q8	oMLX	75.3	3452.0	–	–	65536	2026-10-07	community	–	source
M5 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	75.3	2245.0	–	–	4096	2026-10-11	community	–	source
RTX A6000	gpt-oss 20B 20B	MXFP4	llama.cpp	74.4	1891.2	–	–	32768	–	sourced	–	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	74.4	974.5	–	–	4096	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	74.4	974.5	–	–	4096	2026-10-06	community	contradiction	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	73.8	675.9	–	–	1024	2026-10-11	community	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	73.5	805.6	–	–	1024	2026-10-16	community	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	73.4	612.1	–	–	1024	2026-10-12	community	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	73.4	612.1	–	–	1024	2026-10-12	community	–	source
RTX 5060 Ti 16GB	gpt-oss 20B 20B	MXFP4	llama.cpp	73.2	1737.7	–	–	32768	–	sourced	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	72.6	793.7	–	–	1024	2026-10-13	community	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	72.5	785.6	–	–	1024	2026-10-12	community	–	source
M5 Pro	gpt-oss 20B	OptiQ-4bit	oMLX	72.5	2246.0	–	–	16384	2026-10-14	community	–	source
M5 Pro	gpt-oss 20B	OptiQ-4bit	oMLX	72.5	2246.0	–	–	16384	2026-10-14	community	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	72.2	814.3	–	–	1024	2026-10-09	community	–	source
M5 Pro	gpt-oss 20B	OptiQ-4bit	oMLX	71.5	2272.0	–	–	16384	2026-10-14	community	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	71.2	784.9	–	–	1024	2026-10-13	community	–	source
M1 Max	gpt-oss 20B	4bit	oMLX	71.0	721.9	–	–	4096	2026-10-17	community	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	70.6	775.4	–	–	1024	2026-10-13	community	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	70.4	787.9	–	–	1024	2026-10-16	community	–	source
RTX 4090	gpt-oss 20B 20B	MXFP4	llama.cpp	70.3	1114.7	–	–	131072	–	sourced	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	70.3	808.6	–	–	4096	2026-10-16	community	–	source
M5 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	69.9	2079.0	–	–	1024	2026-10-10	community	contradiction	source
M2 Max	gpt-oss 20B	MXFP4-Q8	oMLX	69.9	622.7	–	–	4096	2026-10-12	community	–	source
M1 Max	gpt-oss 20B	MXFP4-Q8	oMLX	69.9	539.3	–	–	1024	2026-10-17	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	69.8	1373.0	–	–	1024	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	69.8	1373.0	–	–	1024	2026-10-06	community	contradiction	source
M2 Max	gpt-oss 20B	MXFP4-Q8	oMLX	69.7	616.7	–	–	4096	2026-10-17	community	–	source
M2 Max	gpt-oss 20B	MXFP4-Q8	oMLX	69.7	616.7	–	–	4096	2026-10-17	community	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	69.5	662.5	–	–	4096	2026-10-12	community	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	69.5	662.5	–	–	4096	2026-10-12	community	–	source
M1 Max	gpt-oss 20B	MXFP4-Q8	oMLX	69.4	598.7	–	–	1024	2026-10-13	community	–	source
M1 Max	gpt-oss 20B	MXFP4-Q8	oMLX	68.7	623.0	–	–	1024	2026-10-15	community	–	source
M5 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	68.7	2565.0	–	–	16384	2026-10-18	community	–	source
M5 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	68.7	2565.0	–	–	16384	2026-10-18	community	–	source
M1 Max	gpt-oss 20B	MXFP4-Q8	oMLX	68.6	624.1	–	–	1024	2026-10-15	community	–	source
M1 Max	gpt-oss 20B	OptiQ-4bit	oMLX	68.5	569.6	–	–	4096	2026-10-17	community	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	68.2	816.8	–	–	4096	2026-10-09	community	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	67.8	784.0	–	–	4096	2026-10-16	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	67.5	946.8	–	–	4096	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	67.5	946.8	–	–	4096	2026-10-06	community	contradiction	source
M5 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	67.3	2184.0	–	–	1024	2026-10-09	community	contradiction	source
RTX 5070 Ti	gpt-oss 20B 20B	MXFP4	llama.cpp	66.6	1363.7	–	–	131072	–	sourced	–	source
M5 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	66.3	2036.0	–	–	1024	2026-10-10	community	contradiction	source
M5 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	66.2	2627.0	–	–	16384	2026-10-09	community	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	65.9	782.3	–	–	4096	2026-10-11	community	–	source
M1 Max	gpt-oss 20B	MXFP4-Q8	oMLX	65.9	627.0	–	–	4096	2026-10-17	community	–	source
M6	gpt-oss 20B	oQ4	oMLX	65.7	1438.0	–	–	1024	2026-10-08	community	–	source
M1 Max	gpt-oss 20B	MXFP4-Q8	oMLX	64.7	615.7	–	–	4096	2026-10-13	community	–	source
M6	gpt-oss 20B	oQ4	oMLX	64.6	798.1	–	–	1024	2026-10-08	community	–	source
M5 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	64.6	1982.0	–	–	1024	2026-10-10	community	contradiction	source
M1 Max	gpt-oss 20B	4bit	oMLX	64.4	737.0	–	–	8192	2026-10-17	community	–	source
M1 Max	gpt-oss 20B	MXFP4-Q8	oMLX	63.8	609.6	–	–	4096	2026-10-15	community	–	source
M1 Max	gpt-oss 20B	MXFP4-Q8	oMLX	63.7	612.7	–	–	4096	2026-10-15	community	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	63.4	660.8	–	–	1024	2026-10-12	community	contradiction	source
RTX 4060 Ti 16GB	gpt-oss 20B 20B	MXFP4	llama.cpp	63.2	3274.2	–	–	4096	–	sourced	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	63.1	640.2	–	–	8192	2026-10-12	community	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	63.1	640.2	–	–	8192	2026-10-12	community	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	62.5	739.9	–	–	4096	2026-10-12	community	contradiction	source
M2 Max	gpt-oss 20B	MXFP4-Q8	oMLX	62.5	594.5	–	–	16384	2026-10-17	community	–	source
M2 Max	gpt-oss 20B	MXFP4-Q8	oMLX	62.5	594.5	–	–	16384	2026-10-17	community	–	source
M4 Pro	gpt-oss 20B	oQ6e	oMLX	62.4	660.9	–	–	1024	2026-10-12	community	–	source
M4 Pro	gpt-oss 20B	oQ6e	oMLX	62.4	660.9	–	–	1024	2026-10-12	community	–	source
M6	gpt-oss 20B	oQ4	oMLX	62.3	1490.0	–	–	4096	2026-10-08	community	–	source
M1 Max	gpt-oss 20B	OptiQ-4bit	oMLX	61.9	592.8	–	–	8192	2026-10-17	community	–	source
RTX 3090 Ti	gpt-oss 20B 20B	MXFP4	llama.cpp	61.5	936.4	–	–	131072	–	sourced	–	source
M6	gpt-oss 20B	oQ4	oMLX	61.5	1422.0	–	–	4096	2026-10-08	community	–	source
M5 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	61.1	1895.0	–	–	1024	2026-10-10	community	contradiction	source
M4 Pro	gpt-oss 20B	MXFP4-Q4	oMLX	60.9	415.4	–	–	1024	2026-10-15	community	–	source
RTX 4080	gpt-oss 20B 20B	MXFP4	llama.cpp	60.0	898.5	–	–	131072	–	sourced	–	source
RTX 4080 SUPER	gpt-oss 20B 20B	MXFP4	llama.cpp	60.0	1145.0	–	–	131072	–	sourced	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q4	oMLX	59.7	453.3	–	–	1024	2026-10-15	community	–	source
M5 Max	gpt-oss 20B	oQ4	oMLX	59.4	3252.0	–	–	65536	2026-10-09	community	–	source
RTX 5060 Ti 16GB	gpt-oss 20B 20B	MXFP4	llama.cpp	58.1	1102.3	–	–	65536	–	sourced	–	source
RTX 4060 Ti 16GB	gpt-oss 20B 20B	MXFP4	llama.cpp	57.8	2552.9	–	–	16384	–	sourced	–	source
M1 Max	gpt-oss 20B	MXFP4-Q8	oMLX	57.8	595.4	–	–	8192	2026-10-15	community	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	57.7	713.9	–	–	16384	2026-10-13	community	–	source
M1 Max	gpt-oss 20B	MXFP4-Q8	oMLX	57.6	571.1	–	–	8192	2026-10-10	community	–	source
RTX 4070 Ti SUPER	gpt-oss 20B 20B	MXFP4	llama.cpp	57.5	993.5	–	–	131072	–	sourced	–	source
RTX 6000 Ada	gpt-oss 20B 20B	MXFP4	llama.cpp	57.5	1177.7	–	–	131072	–	sourced	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	57.4	714.7	–	–	16384	2026-10-13	community	–	source
M5 Pro	gpt-oss 20B	OptiQ-4bit	oMLX	57.4	1913.0	–	–	32768	2026-10-14	community	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	56.8	712.0	–	–	16384	2026-10-12	community	–	source
M1 Max	gpt-oss 20B	4bit	oMLX	56.0	693.7	–	–	16384	2026-10-17	community	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	55.9	589.3	–	–	16384	2026-10-12	community	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	55.9	589.3	–	–	16384	2026-10-12	community	–	source
Arc Pro B70	gpt-oss 20B	MXFP4	llama.cpp (SYCL)	55.84	–	–	–	–	2026-10-01	community	–	source
M2 Max	gpt-oss 20B	MXFP4-Q8	oMLX	55.4	564.7	–	–	16384	2026-10-12	community	contradiction	source
M2 Max	gpt-oss 20B	MXFP4-Q8	oMLX	54.5	567.0	–	–	16384	2026-10-08	community	contradiction	source
M4 Pro	gpt-oss 20B	unknown	oMLX	54.5	621.4	–	–	1024	2026-10-11	community	–	source
M5 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	53.4	2027.0	–	–	32768	2026-10-15	community	–	source
RTX A6000	gpt-oss 20B 20B	MXFP4	llama.cpp	53.3	1207.2	–	–	65536	–	sourced	–	source
M5 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	53.3	2001.0	–	–	32768	2026-10-15	community	–	source
M1 Max	gpt-oss 20B	OptiQ-4bit	oMLX	53.2	554.6	–	–	16384	2026-10-17	community	–	source
M3 Pro	gpt-oss 20B	OptiQ-4bit	oMLX	52.8	642.4	–	–	1024	2026-10-12	community	–	source
M4 Pro	gpt-oss 20B	8bit	oMLX	51.9	562.4	–	–	1024	2026-10-06	community	–	source
M4 Pro	gpt-oss 20B	8bit	oMLX	51.9	562.4	–	–	1024	2026-10-06	community	–	source
RTX 4060 Ti 16GB	gpt-oss 20B 20B	MXFP4	llama.cpp	51.5	1964.9	–	–	32768	–	sourced	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q4	oMLX	51.1	454.4	–	–	8192	2026-10-15	community	–	source
M3 Pro	gpt-oss 20B	OptiQ-4bit	oMLX	50.8	635.6	–	–	4096	2026-10-12	community	–	source
M1 Max	gpt-oss 20B	MXFP4-Q8	oMLX	50.8	553.1	–	–	16384	2026-10-15	community	–	source
M4 Pro	gpt-oss 20B	8bit	oMLX	49.7	626.9	–	–	4096	2026-10-06	community	–	source
M4 Pro	gpt-oss 20B	8bit	oMLX	49.7	626.9	–	–	4096	2026-10-06	community	–	source
M6	gpt-oss 20B	MXFP4-Q8	oMLX	49.6	1657.0	–	–	1024	2026-10-08	community	–	source
M6	gpt-oss 20B	MXFP4-Q8	oMLX	48.7	1654.0	–	–	1024	2026-10-08	community	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	48.6	621.6	–	–	32768	2026-10-12	community	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	48.1	614.9	–	–	32768	2026-10-12	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	48.0	454.5	–	–	1024	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	48.0	454.5	–	–	1024	2026-10-06	community	contradiction	source
M6	gpt-oss 20B	MXFP4-Q8	oMLX	47.5	1879.0	–	–	4096	2026-10-08	community	–	source
M6	gpt-oss 20B	MXFP4-Q8	oMLX	47.4	1887.0	–	–	4096	2026-10-08	community	–	source
M6	gpt-oss 20B	MXFP4-Q8	oMLX	47.3	681.9	–	–	1024	2026-10-08	community	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	45.8	504.9	–	–	32768	2026-10-12	community	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	45.8	504.9	–	–	32768	2026-10-12	community	–	source
M6	gpt-oss 20B	MXFP4-Q8	oMLX	45.4	1848.0	–	–	8192	2026-10-08	community	–	source
M5 Max	gpt-oss 20B	MXFP4-Q8	oMLX	45.2	1657.0	–	–	65536	2026-10-14	community	contradiction	source
M1 Max	gpt-oss 20B	4bit	oMLX	45.2	598.6	–	–	32768	2026-10-17	community	–	source
M6	gpt-oss 20B	MXFP4-Q8	oMLX	45.1	1769.0	–	–	4096	2026-10-08	community	–	source
M6	gpt-oss 20B 21B-A3.6B	unstated	LM Studio	45	–	–	–	–	2026-10	community	–	source
M4 Pro	gpt-oss 20B	oQ6e	oMLX	44.2	601.7	–	–	32768	2026-10-12	community	–	source
M4 Pro	gpt-oss 20B	oQ6e	oMLX	44.2	601.7	–	–	32768	2026-10-12	community	–	source
RTX 5060 Ti 16GB	gpt-oss 20B 20B	MXFP4	llama.cpp	43.8	685.3	–	–	131072	–	sourced	–	source
M1 Max	gpt-oss 20B	OptiQ-4bit	oMLX	41.9	490.2	–	–	32768	2026-10-17	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	41.4	880.1	–	–	1024	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	41.4	880.1	–	–	1024	2026-10-06	community	contradiction	source
M3 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	41.2	664.2	–	–	4096	2026-10-15	community	–	source
M3 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	41.2	664.2	–	–	4096	2026-10-15	community	–	source
RTX 4060 Ti 16GB	gpt-oss 20B 20B	MXFP4	llama.cpp	41.1	1332.2	–	–	65536	–	sourced	–	source
M5 Pro	gpt-oss 20B	OptiQ-4bit	oMLX	41.1	1536.0	–	–	65536	2026-10-14	community	–	source
M1 Max	gpt-oss 20B	MXFP4-Q8	oMLX	40.5	489.8	–	–	32768	2026-10-15	community	–	source
M1 Max	gpt-oss 20B	MXFP4-Q8	oMLX	40.4	492.0	–	–	32768	2026-10-15	community	–	source
M5 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	38.7	1556.0	–	–	65536	2026-10-15	community	–	source
M5 Max	gpt-oss 20B	oQ4	oMLX	38.6	2302.0	–	–	131072	2026-10-09	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	38.4	878.3	–	–	4096	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	38.4	878.3	–	–	4096	2026-10-06	community	contradiction	source
M5 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	37.2	1502.0	–	–	65536	2026-10-15	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	36.6	944.9	–	–	4096	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	36.6	944.9	–	–	4096	2026-10-06	community	contradiction	source
M3 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	36.5	584.5	–	–	16384	2026-10-15	community	–	source
M3 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	36.5	584.5	–	–	16384	2026-10-15	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	36.3	862.0	–	–	1024	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	36.3	862.0	–	–	1024	2026-10-06	community	contradiction	source
M5 Max	gpt-oss 20B	MXFP4-Q8	oMLX	36.3	1584.0	–	–	199680	2026-10-07	community	–	source
M4	gpt-oss 20B	MXFP4-Q8	oMLX	34.9	373.6	–	–	1024	2026-10-18	community	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	34.6	459.3	–	–	65536	2026-10-12	community	–	source
M4	gpt-oss 20B	MXFP4-Q8	oMLX	34.2	339.9	–	–	4096	2026-10-18	community	–	source
M4	gpt-oss 20B	MXFP4-Q8	oMLX	34.2	349.6	–	–	4096	2026-10-18	community	–	source
M4	gpt-oss 20B	MXFP4-Q8	oMLX	34.0	320.3	–	–	1024	2026-10-12	community	–	source
M4	gpt-oss 20B	MXFP4-Q8	oMLX	34.0	320.3	–	–	1024	2026-10-12	community	–	source
RTX A6000	gpt-oss 20B 20B	MXFP4	llama.cpp	33.4	702.8	–	–	131072	–	sourced	–	source
M4	gpt-oss 20B	MXFP4-Q8	oMLX	32.6	324.9	–	–	8192	2026-10-18	community	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	32.5	397.5	–	–	65536	2026-10-12	community	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	32.5	397.5	–	–	65536	2026-10-12	community	–	source
M4	gpt-oss 20B	MXFP4-Q8	oMLX	32.1	336.0	–	–	4096	2026-10-12	community	–	source
M4	gpt-oss 20B	MXFP4-Q8	oMLX	32.1	336.0	–	–	4096	2026-10-12	community	–	source
M1 Max	gpt-oss 20B	4bit	oMLX	32.1	469.9	–	–	65536	2026-10-17	community	–	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	32.0	354.6	–	–	4096	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q8	oMLX	32.0	354.6	–	–	4096	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	31.5	1027.0	–	–	4096	2026-10-06	community	contradiction	source
M3 Max	gpt-oss 20B	MXFP4-Q4LH	oMLX	31.5	1027.0	–	–	4096	2026-10-06	community	contradiction	source
RTX 4060 Ti 16GB	gpt-oss 20B 20B	MXFP4	llama.cpp	31.1	780.2	–	–	131072	–	sourced	–	source
M3 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	30.8	502.3	–	–	32768	2026-10-15	community	–	source
M3 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	30.8	502.3	–	–	32768	2026-10-15	community	–	source
M5 Max	gpt-oss 20B	MXFP4-Q8	oMLX	30.0	1131.0	–	–	131072	2026-10-14	community	–	source
M1 Max	gpt-oss 20B	OptiQ-4bit	oMLX	29.0	394.2	–	–	65536	2026-10-17	community	–	source
M1 Max	gpt-oss 20B	MXFP4-Q8	oMLX	28.2	400.5	–	–	65536	2026-10-15	community	–	source
M1 Max	gpt-oss 20B	MXFP4-Q8	oMLX	28.2	399.2	–	–	65536	2026-10-15	community	–	source
M4	gpt-oss 20B	MXFP4-Q8	oMLX	27.8	265.3	–	–	16384	2026-10-18	community	–	source
M4	gpt-oss 20B	MXFP4-Q4	oMLX	27.4	311.7	–	–	32768	2026-10-13	community	–	source
M3 Pro	gpt-oss 20B	OptiQ-4bit	oMLX	26.1	526.4	–	–	4096	2026-10-13	community	contradiction	source
M5 Pro	gpt-oss 20B	OptiQ-4bit	oMLX	25.7	1145.0	–	–	131072	2026-10-14	community	–	source
M4	gpt-oss 20B	OptiQ-4bit	oMLX	25.6	309.3	–	–	32768	2026-10-13	community	–	source
M5 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	24.8	1143.0	–	–	131072	2026-10-15	community	–	source
M5 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	24.7	1126.0	–	–	131072	2026-10-15	community	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	24.3	339.7	–	–	131072	2026-10-13	community	–	source
M3 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	23.1	390.2	–	–	65536	2026-10-15	community	–	source
M3 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	23.1	390.2	–	–	65536	2026-10-15	community	–	source
M4	gpt-oss 20B	oQ5e	oMLX	22.4	301.8	–	–	32768	2026-10-15	community	–	source
M4 Max (128 GB)	gpt-oss 20B	MXFP4	llama.cpp	21.8	–	–	–	32768	2026-10-17	sourced	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	21.1	276.5	–	–	131072	2026-10-12	community	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	21.1	276.5	–	–	131072	2026-10-12	community	–	source
M1 Max	gpt-oss 20B	4bit	oMLX	19.7	328.4	–	–	131072	2026-10-17	community	–	source
M5 Pro	gpt-oss 20B	OptiQ-4bit	oMLX	18.3	868.6	–	–	199680	2026-10-14	community	–	source
M1 Max	gpt-oss 20B	OptiQ-4bit	oMLX	18.0	283.6	–	–	131072	2026-10-17	community	–	source
M3 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	17.0	266.8	–	–	131072	2026-10-15	community	–	source
M4	gpt-oss 20B	oQ5e	oMLX	16.3	233.2	–	–	65536	2026-10-16	community	–	source
M4 Max (128 GB)	gpt-oss 20B	MXFP4	llama.cpp	15.6	–	–	–	102400	2026-10-17	sourced	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	15.4	208.7	–	–	199680	2026-10-12	community	–	source
M4 Pro	gpt-oss 20B	MXFP4-Q8	oMLX	15.4	208.7	–	–	199680	2026-10-12	community	–	source
M3 Pro	gpt-oss 20B	OptiQ-4bit	oMLX	13.8	619.3	–	–	1024	2026-10-13	community	contradiction	source
3x RTX PRO 4000 Blackwell	gpt-oss 20B	unspecified	CUDA	–	17812	–	–	–	2026-10	community	–	source
3x RTX PRO 4000 Blackwell	gpt-oss 20B	unspecified	CUDA	–	17746	–	–	–	2026-10	community	–	source

← all models