OpenAI
OpenAI
OpenAI/gpt-oss-20b

Runs

View all benchmark runs for this model family.

	Quant						Actions

	Quant		Actions

	Quant						Actions
M4 Pro	q5_k_m	llama.cpp0.3.0	64.9 tok/s	635.2 tok/s	6.21 GB 26%
M2 Max	mxfp4	mlx_lm0.31.3	57.8 tok/s	465.8 tok/s	14.00 GB 44%
M2 Max	mxfp4	mlx_lm0.31.3	66.4 tok/s	493.3 tok/s	13.00 GB 41%
M5 Pro	mxfp4	mlx_lm0.31.3	79.7 tok/s	2,884.9 tok/s	13.00 GB 20%
M4 Pro	mxfp4	mlx_lm0.31.3	81.7 tok/s	671.3 tok/s	13.00 GB 54%
M4 Max	4bit	mlx_lm0.31.2	91.3 tok/s	1,307.2 tok/s	18.00 GB 14%
M5 Max	mxfp4	mlx_lm0.31.2	114.9 tok/s	4,184.3 tok/s	14.00 GB 22%
M4	mxfp4	mlx_lm0.31.2	25.0 tok/s	249.9 tok/s	12.00 GB 38%
M4	mxfp4	mlx_lm0.31.2	25.7 tok/s	225.6 tok/s	12.00 GB 38%
M3 Max	mxfp4	mlx_lm0.30.2	25.9 tok/s	740.5 tok/s	12.00 GB 9%

	Quant		Actions
M4 Pro	q5_k_m	64.9 tok/s
M2 Max	mxfp4	57.8 tok/s
M2 Max	mxfp4	66.4 tok/s
M5 Pro	mxfp4	79.7 tok/s
M4 Pro	mxfp4	81.7 tok/s
M4 Max	4bit	91.3 tok/s
M5 Max	mxfp4	114.9 tok/s
M4	mxfp4	25.0 tok/s
M4	mxfp4	25.7 tok/s
M3 Max	mxfp4	25.9 tok/s