← Back to community benchmarks

Qwen3.6-35B-A3B-Abliterated-Heretic

M5 Max (40c) · 128 GB · 8bit · 2026-04-28
Performance
16k
tokens
3,657
PP tok/s
83.5
TG tok/s
4481
TTFT (ms)
38.1
Peak mem (GB)
Hardware
Chip M5 Max (40c)
Memory 128 GB
GPU Cores 40
Software
oMLX v0.3.8.dev3
macOS macOS 26.4.1
Context 16,384
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 2,000 92.9 36.4 GB view
4k 3,492 93.2 37.1 GB view
8k 3,732 89.6 37.5 GB view
16k 3,657 83.5 38.1 GB current
32k 3,123 73.5 39.5 GB view
64k 2,178 54.0 42.1 GB view
Batching Results
Batch Size TG tok/s Speedup
92.9 1.00×
153.6 1.65×
222.3 2.39×