← Back to community benchmarks

MiniMax-M2.1

M3 Ultra (80c) · 512 GB · 4bit · 2026-03-10
Performance
32k
tokens
438.1
PP tok/s
17.6
TG tok/s
74803
TTFT (ms)
128.5
Peak mem (GB)
Hardware
Chip M3 Ultra (80c)
Memory 512 GB
GPU Cores 80
Software
oMLX v0.2.6
macOS macOS 26.3
Context 32,768
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 664.5 47.0 120.7 GB view
4k 752.6 40.5 121.7 GB view
8k 696.4 33.7 122.6 GB view
16k 589.0 26.7 124.6 GB view
32k 438.1 17.6 128.5 GB current
64k 296.7 11.6 136.4 GB view
Batching Results
Batch Size TG tok/s Speedup
47.0 1.00×
71.1 1.51×
101.0 2.15×
137.0 2.91×