Google
Google
Google/Gemma 4 E2B IT

Runs

View all benchmark runs for this model family.

	Quant						Actions

	Quant		Actions

	Quant						Actions
M3 Pro	q4_k_m	llama.cppb9430	62.5 tok/s	1,071.9 tok/s	3.95 GB 11%
M4	iq4_nl	llama.cppb8680	44.5 tok/s	664.0 tok/s	0.86 GB 4%
Arc Graphics	ud_q4_k_xl	llama.cppb9010	8.7 tok/s	116.9 tok/s	4.97 GB 16%
M3 Pro	q4_k_m	llama.cppb9010	57.7 tok/s	967.0 tok/s	3.21 GB 9%
M4	4bit	mlx_lm0.31.2	52.0 tok/s	1,405.9 tok/s	3.74 GB 12%
M5	8bit	mlx_lm0.31.2	44.3 tok/s	5,519.6 tok/s	6.69 GB 21%
M5	4bit	mlx_lm0.31.2	74.8 tok/s	5,728.3 tok/s	4.39 GB 14%
M3 Pro	q4_k_m	llama.cppb8640	60.9 tok/s	1,059.4 tok/s	2.87 GB 8%
M1 Max	OptiQ-4bit	mlx_lm0.31.2	75.3 tok/s	2,009.4 tok/s	5.42 GB 8%
M1 Max	8bit	mlx_lm0.31.2	68.1 tok/s	2,159.2 tok/s	6.17 GB 10%

	Quant		Actions
M3 Pro	q4_k_m	62.5 tok/s
M4	iq4_nl	44.5 tok/s
Arc Graphics	ud_q4_k_xl	8.7 tok/s
M3 Pro	q4_k_m	57.7 tok/s
M4	4bit	52.0 tok/s
M5	8bit	44.3 tok/s
M5	4bit	74.8 tok/s
M3 Pro	q4_k_m	60.9 tok/s
M1 Max	OptiQ-4bit	75.3 tok/s
M1 Max	8bit	68.1 tok/s