← Back to community benchmarks

NVIDIA-Nemotron-3-Super-120B-A12B-OptiQ

M5 Max (40c) · 128 GB · 4bit · 2026-07-17
Performance
16k
tokens
415.1
PP tok/s
24.6
TG tok/s
39468
TTFT (ms)
51.8
Peak mem (GB)
Hardware
Chip M5 Max (40c)
Memory 128 GB
GPU Cores 40
Software
oMLX v0.5.1
macOS macOS 27.0
Context 16,384
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 349.4 25.7 47.9 GB view
4k 410.9 25.4 51.0 GB view
8k 418.9 25.1 51.4 GB view
16k 415.1 24.6 51.8 GB current
32k 401.0 23.5 52.9 GB view
64k 395.3 21.4 52.6 GB view
Batching Results
Batch Size TG tok/s Speedup
25.7 1.00×
35.0 1.36×
49.2 1.91×