← Back to community benchmarks

Qwen3.5-35B-A3B

M3 Ultra (80c) · 512 GB · 4bit · 2026-03-13
Performance
8k
tokens
2,642
PP tok/s
84.7
TG tok/s
3100
TTFT (ms)
20.4
Peak mem (GB)
Hardware
Chip M3 Ultra (80c)
Memory 512 GB
GPU Cores 80
Software
oMLX v0.2.10
macOS macOS 26.3.1
Context 8,192
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 1,843 90.4 20.1 GB view
4k 2,435 85.2 20.2 GB view
8k 2,642 84.7 20.4 GB current
Batching Results
Batch Size TG tok/s Speedup
90.4 1.00×
158.8 1.76×
259.8 2.87×