← Back to community benchmarks

Kimi-K2.6DQ3_K_M-q8

M3 Ultra (80c) · 512 GB · 4bit · 2026-04-21
Performance
16k
tokens
188.4
PP tok/s
16.3
TG tok/s
86955
TTFT (ms)
448.2
Peak mem (GB)
Hardware
Chip M3 Ultra (80c)
Memory 512 GB
GPU Cores 80
Software
oMLX v0.3.6
macOS macOS 26.4.1
Context 16,384
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 211.1 20.3 439.6 GB view
4k 238.1 20.0 441.2 GB view
8k 219.5 18.6 443.4 GB view
16k 188.4 16.3 448.2 GB current
32k 121.7 13.6 457.8 GB view
Batching Results
Batch Size TG tok/s Speedup
20.3 1.00×
30.9 1.52×
38.6 1.90×
47.9 2.36×