← Back to community benchmarks

Nex-N2-Pro5.3bit-vision

M3 Ultra (80c) · 512 GB · 4bit · 2026-07-13
Performance
16k
tokens
509.3
PP tok/s
27.9
TG tok/s
32168
TTFT (ms)
251.1
Peak mem (GB)
Hardware
Chip M3 Ultra (80c)
Memory 512 GB
GPU Cores 80
Software
oMLX v0.5.1
macOS macOS 26.5.2
Context 16,384
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 482.6 30.0 248.7 GB view
4k 546.6 29.6 250.3 GB view
8k 536.4 29.0 250.8 GB view
16k 509.3 27.9 251.1 GB current
32k 459.3 26.6 251.6 GB view
64k 383.6 24.2 253.2 GB view
128k 287.7 20.7 255.0 GB view
195k 223.4 17.8 257.1 GB view
Batching Results
Batch Size TG tok/s Speedup
30.0 1.00×
46.4 1.55×
69.9 2.33×