Gemma 4 Apple Silicon Benchmarks

E2B and E4B are effective-parameter mixtures. The 12B is dense. 26B-A4B is a mixture-of-experts. GroundSlate runs the 4-bit MLX conversions from mlx-community.

ai.google.dev/gemma

Models

Google DeepMind · Gemma 4

Gemma 4 E2B

mlx-community/gemma-4-e2b-it-4bit

3.0 GB

Google DeepMind · Gemma 4

Gemma 4 E4B

mlx-community/gemma-4-e4b-it-4bit

5.2 GB

Google DeepMind · Gemma 4

Gemma 4 12B

mlx-community/gemma-4-12B-it-4bit

8.1 GB

Google DeepMind · Gemma 4

Gemma 4 26B-A4B

mlx-community/gemma-4-26b-a4b-it-4bit

15.5 GB