← Back to community benchmarks

Qwen3.5-2B

M5 Max (40c) · 128 GB · 4bit · 2026-04-02
User comment2k,4k,8k,16k not reported due to https://github.com/jundot/omlx/issues/524
Performance
1k
tokens
3,438
PP tok/s
326.7
TG tok/s
298
TTFT (ms)
2.1
Peak mem (GB)
Hardware
Chip M5 Max (40c)
Memory 128 GB
GPU Cores 40
Software
oMLX v0.3.0
macOS macOS 26.4
Context 1,024
Batching Results
Batch Size TG tok/s Speedup
326.7 1.00×
574.8 1.76×
824.8 2.52×
967.2 2.96×