← Back to community benchmarks

Qwen3.5-122B-A10B-oQ3

M1 Max (32c) · 64 GB · 3bit · 2026-07-21
Performance
64k
tokens
104.2
PP tok/s
19.8
TG tok/s
628735
TTFT (ms)
52.3
Peak mem (GB)
Hardware
Chip M1 Max (32c)
Memory 64 GB
GPU Cores 32
Software
oMLX v0.5.2rc1
macOS macOS 26.5.2
Context 65,536
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 177.5 32.2 47.9 GB view
16k 163.5 27.8 50.6 GB view
64k 104.2 19.8 52.3 GB current
Batching Results
Batch Size TG tok/s Speedup
32.2 1.00×
38.5 1.20×
76.8 2.39×