← Back to community benchmarks

gemma-3-4b-it-qat

M1 Max (24c) · 32 GB · 4bit · 2026-03-14
Performance
8k
tokens
484.0
PP tok/s
64.5
TG tok/s
16929
TTFT (ms)
4.0
Peak mem (GB)
Hardware
Chip M1 Max (24c)
Memory 32 GB
GPU Cores 24
Software
oMLX v0.2.12
macOS macOS 26.2
Context 8,192
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 493.5 73.4 3.6 GB view
4k 487.6 69.5 3.8 GB view
8k 484.0 64.5 4.0 GB current
16k 477.5 54.7 4.3 GB view
32k 462.8 38.8 4.8 GB view
64k 431.3 18.6 6.0 GB view
Batching Results
Batch Size TG tok/s Speedup
73.4 1.00×
88.3 1.20×
93.3 1.27×