Google
Google
/Gemma 4 31B

Quantizations

QuantQuantized bySizeDecodePrefillScoreActions
MLX Community
MLX Community
16.3 GB14.6 tok/s364.6 tok/sRuns poorly
MLX Community
MLX Community
31.4 GB7.7 tok/s305.6 tok/sBarely runs

Device Comparison

Results include trials with 4,096 input tokens and 1,024 output tokens only.

Decode / Prefill Speeds

5 devices