← Back to community benchmarks

Hermes-2-Pro-Llama-3-8B

M3 Ultra (80c) · 512 GB · unknown · 2026-03-13
Performance
1k
tokens
1,191
PP tok/s
40.5
TG tok/s
861
TTFT (ms)
15.4
Peak mem (GB)
Hardware
Chip M3 Ultra (80c)
Memory 512 GB
GPU Cores 80
Software
oMLX v0.2.10
macOS macOS 26.3.1
Context 1,024
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 1,191 40.5 15.4 GB current
4k 1,452 39.2 16.0 GB view
Batching Results
Batch Size TG tok/s Speedup
40.5 1.00×
46.9 1.16×
91.4 2.26×