← Back to community benchmarks

Qwen3.5-35B-A3B-5bit

M3 Ultra (60c) · 256 GB · 5bit · 2026-03-10
Performance
32k
tokens
1,636
PP tok/s
64.0
TG tok/s
20035
TTFT (ms)
25.7
Peak mem (GB)
Hardware
Chip M3 Ultra (60c)
Memory 256 GB
GPU Cores 60
Software
oMLX v0.2.7
macOS macOS 26.3
Context 32,768
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 2,010 87.8 24.1 GB view
4k 2,093 85.0 24.2 GB view
8k 2,030 80.4 24.4 GB view
16k 1,887 73.6 24.9 GB view
32k 1,636 64.0 25.7 GB current
64k 1,271 49.3 27.4 GB view
Batching Results
Batch Size TG tok/s Speedup
87.8 1.00×
144.0 1.64×
212.6 2.42×
279.2 3.18×