← Back to community benchmarks

Llama-3.3-70B-Instruct

M2 Max (38c) · 96 GB · 4bit · 2026-03-10

Performance

4k

tokens

64.7

PP tok/s

7.7

TG tok/s

63303

TTFT (ms)

39.0

Peak mem (GB)

Hardware

Chip M2 Max (38c)

Memory 96 GB

GPU Cores 38

Software

oMLX v0.2.6

macOS macOS 26.2

Context 4,096

Performance by Context Length

Context	PP tok/s	TG tok/s	Peak Mem
1k	72.7	8.6	37.8 GB	view
4k	64.7	7.7	39.0 GB	current

Batching Results

Batch Size	TG tok/s	Speedup
1×	8.6	1.00×
2×	12.4	1.44×
4×	10.4	1.21×