← Back to community benchmarks

MiniMax-M2.5

M3 Ultra (80c) · 512 GB · 8bit · 2026-03-09
Performance
16k
tokens
570.5
PP tok/s
21.8
TG tok/s
28719
TTFT (ms)
231.1
Peak mem (GB)
Hardware
Chip M3 Ultra (80c)
Memory 512 GB
GPU Cores 80
Software
oMLX v0.2.6
macOS macOS 26.3
Context 16,384
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 564.2 33.2 227.2 GB view
4k 702.1 30.1 228.2 GB view
8k 666.0 26.9 229.1 GB view
16k 570.5 21.8 231.1 GB current
32k 429.4 15.6 235.0 GB view
64k 284.2 10.9 242.8 GB view
Batching Results
Batch Size TG tok/s Speedup
33.2 1.00×
49.7 1.50×
63.1 1.90×