← Back to community benchmarks

Qwen3-Coder-Next

M3 Ultra (60c) · 96 GB · 4bit · 2026-04-07

Performance

1k

tokens

1,346

PP tok/s

69.6

TG tok/s

761

TTFT (ms)

43.1

Peak mem (GB)

Hardware

Chip M3 Ultra (60c)

Memory 96 GB

GPU Cores 60

Software

oMLX v0.3.4

macOS macOS 26.3.1

Context 1,024

Performance by Context Length

Context	PP tok/s	TG tok/s	Peak Mem
1k	1,346	69.6	43.1 GB	current
4k	1,653	68.8	44.0 GB	view

Batching Results

Batch Size	TG tok/s	Speedup
1×	69.6	1.00×
2×	117.1	1.68×
4×	185.1	2.66×