← Back to community benchmarks

Qwen3.5-35B-A3B

M3 Ultra (80c) · 512 GB · 4bit · 2026-03-15
Performance
4k
tokens
2,751
PP tok/s
93.2
TG tok/s
1489
TTFT (ms)
20.2
Peak mem (GB)
Hardware
Chip M3 Ultra (80c)
Memory 512 GB
GPU Cores 80
Software
oMLX v0.2.12
macOS macOS 26.3.1
Context 4,096
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 2,638 96.7 20.1 GB view
4k 2,751 93.2 20.2 GB current
8k 2,667 82.7 20.4 GB view
16k 2,468 77.1 20.8 GB view
Batching Results
Batch Size TG tok/s Speedup
96.7 1.00×
171.9 1.78×
281.0 2.91×