← Back to community benchmarks

Qwen3.6-35B-A3B

M3 Ultra (60c) · 96 GB · 4bit · 2026-04-19
Performance
64k
tokens
1,315
PP tok/s
48.9
TG tok/s
49854
TTFT (ms)
25.0
Peak mem (GB)
Hardware
Chip M3 Ultra (60c)
Memory 96 GB
GPU Cores 60
Software
oMLX v0.3.6
macOS macOS 26.4
Context 65,536
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 1,491 103.6 19.3 GB view
4k 2,098 100.6 20.0 GB view
8k 2,131 93.0 20.4 GB view
16k 2,017 83.5 21.0 GB view
32k 1,745 70.4 22.4 GB view
64k 1,315 48.9 25.0 GB current
128k 838.1 26.7 30.4 GB view
195k 606.9 14.5 36.1 GB view
Batching Results
Batch Size TG tok/s Speedup
103.6 1.00×
174.5 1.68×
261.9 2.53×
378.2 3.65×