← Back to community benchmarks

Qwen3.5-35B-A3B-oQ4e

M1 Ultra (48c) · 64 GB · 4bit · 2026-04-17
Performance
32k
tokens
2,098
PP tok/s
35.1
TG tok/s
15616
TTFT (ms)
23.2
Peak mem (GB)
Hardware
Chip M1 Ultra (48c)
Memory 64 GB
GPU Cores 48
Software
oMLX v0.3.6
macOS macOS 26.3.1
Context 32,768
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 594.2 58.8 20.4 GB view
4k 838.4 54.5 21.1 GB view
8k 851.8 53.1 21.5 GB view
16k 2,238 36.9 22.6 GB view
32k 2,098 35.1 23.2 GB current
64k 1,739 34.7 25.1 GB view
128k 1,123 33.3 27.8 GB view
195k 792.6 31.0 30.7 GB view
Batching Results
Batch Size TG tok/s Speedup
58.8 1.00×
96.7 1.64×
137.8 2.34×
179.2 3.05×