← Back to community benchmarks

Qwen3-4B-Instruct-2507-oQ4

M4 (10c) · 16 GB · 4bit · 2026-04-05
Performance
1k
tokens
384.3
PP tok/s
34.9
TG tok/s
2664
TTFT (ms)
2.9
Peak mem (GB)
Hardware
Chip M4 (10c)
Memory 16 GB
GPU Cores 10
Software
oMLX v0.3.3
macOS macOS 26.4
Context 1,024
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 384.3 34.9 2.9 GB current
4k 362.6 27.5 3.3 GB view
Batching Results
Batch Size TG tok/s Speedup
34.9 1.00×
60.9 1.74×
78.1 2.24×