← Back to community benchmarks

Qwen2.5-Coder-1.5B-Instruct

M3 Ultra (80c) · 512 GB · 4bit · 2026-04-30

Performance

1k

tokens

3,699

PP tok/s

261.4

TG tok/s

277

TTFT (ms)

1.5

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 1,024

Batching Results

Batch Size	TG tok/s	Speedup
1×	261.4	1.00×
2×	475.3	1.82×
4×	749.9	2.87×