Google
Google
Google/Gemma 4 26B A4B IT

Runs

View all benchmark runs for this model family.

	Quant
M2 Max	4bit	mlx_lm0.31.2	50.6 tok/s	446.1 tok/s	16.00 GB 17%
M4 Max	4bit	mlx_lm0.31.2	76.2 tok/s	1,195.2 tok/s	17.00 GB 47%
M4 Max	ud-q4_k_m	llama.cppb8680	67.9 tok/s	1,066.1 tok/s	16.26 GB 45%
M1 Max	ud-iq3_xxs	llama.cppb8640	38.4 tok/s	411.0 tok/s	11.39 GB 18%
M2 Pro	4bit	mlx_lm0.31.2	44.0 tok/s	331.4 tok/s	16.00 GB 50%
M4 Max	ud-q4_k_m	llama.cppb8670	70.1 tok/s	1,090.5 tok/s	9.26 GB 14%
RTX PRO 4000 Blackwell	ud-q4_k_m	llama.cppb8672	103.2 tok/s	3,055.9 tok/s	16.24 GB 26%
M4 Max	q4_k_m	llama.cppb8660	83.8 tok/s	1,074.0 tok/s	10.71 GB 8%
M4 Pro	q4_k_m	llama.cppb8793	57.0 tok/s	700.5 tok/s	13.16 GB 21%
M4 Pro	ud-q4_k_xl	llama.cppb8793	48.5 tok/s	705.5 tok/s	15.95 GB 25%