← Back to community benchmarks

Qwen3.5-122B-A10B6.5bit

M3 Ultra (80c) · 256 GB · 6bit · 2026-04-03
Performance
8k
tokens
913.5
PP tok/s
45.2
TG tok/s
8968
TTFT (ms)
96.1
Peak mem (GB)
Hardware
Chip M3 Ultra (80c)
Memory 256 GB
GPU Cores 80
Software
oMLX v0.3.2
macOS macOS 26.4
Context 8,192
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 718.8 49.4 94.0 GB view
4k 903.1 47.6 95.5 GB view
8k 913.5 45.2 96.1 GB current
16k 864.1 41.0 97.3 GB view
Batching Results
Batch Size TG tok/s Speedup
49.4 1.00×
78.7 1.59×
105.2 2.13×
137.6 2.79×