← Back to community benchmarks

Qwen3-4B-Instruct-2507

M5 Max (40c) · 128 GB · bf16 · 2026-05-04
Performance
1k
tokens
3,672
PP tok/s
56.9
TG tok/s
279
TTFT (ms)
8.2
Peak mem (GB)
Hardware
Chip M5 Max (40c)
Memory 128 GB
GPU Cores 40
Software
oMLX v0.3.8
macOS macOS 26.4.1
Context 1,024
Batching Results
Batch Size TG tok/s Speedup
56.9 1.00×
57.3 1.01×
105.6 1.86×
198.1 3.48×