← Back to community benchmarks

Qwen3.5-122B-A10B

M3 Ultra (80c) · 512 GB · 4bit · 2026-03-16
Performance
32k
tokens
738.1
PP tok/s
36.3
TG tok/s
44395
TTFT (ms)
69.1
Peak mem (GB)
Hardware
Chip M3 Ultra (80c)
Memory 512 GB
GPU Cores 80
Software
oMLX v0.2.10
macOS macOS 15.7.4
Context 32,768
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 891.2 52.4 66.3 GB view
4k 935.4 50.3 66.7 GB view
8k 914.1 46.7 67.0 GB view
16k 855.6 43.1 67.7 GB view
32k 738.1 36.3 69.1 GB current
Batching Results
Batch Size TG tok/s Speedup
52.4 1.00×
86.1 1.64×
130.5 2.49×
187.3 3.57×