← Back to community benchmarks
Qwen3.5-9B-DeepSeek-V4-Flash-oQ4-mtp
Performance
1k
tokens
176.3
PP tok/s
27.3
TG tok/s
5809
TTFT (ms)
6.0
Peak mem (GB)
Hardware
Chip
M4 (10c)
Memory
24 GB
GPU Cores
10
Software
oMLX
v0.3.9.dev2
macOS
macOS 26.4.1
Context
1,024