← Back to community benchmarks

MiniMax-M2.5-CARVE-v1-Q8

M3 Ultra (80c) · 512 GB · 8bit · 2026-03-16
Performance
16k
tokens
573.8
PP tok/s
22.1
TG tok/s
28556
TTFT (ms)
231.1
Peak mem (GB)
Hardware
Chip M3 Ultra (80c)
Memory 512 GB
GPU Cores 80
Software
oMLX v0.2.14
macOS macOS 26.3
Context 16,384
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 619.8 35.4 227.2 GB view
4k 721.1 31.4 228.2 GB view
8k 665.2 26.4 229.1 GB view
16k 573.8 22.1 231.1 GB current
Batching Results
Batch Size TG tok/s Speedup
35.4 1.00×
49.9 1.41×
65.5 1.85×
100.2 2.83×