Community Benchmarks

Download oMLX →

Real performance numbers from oMLX users around the world. Submit yours from the app (v0.2.6+).

# Chip RAM Model Quant Ctx PP tok/s TG tok/s Date ↓
2051 M4 (10c) 16 GB Qwen3.5-9B 4bit 4k 220.1 18.8 26-03-09
2052 M2 Pro (16c) 16 GB Qwen3.5-9B 4bit 1k 162.6 34.0 26-03-09
2053 M2 Pro (16c) 16 GB Qwen3.5-9B 4bit 4k 173.8 34.0 26-03-09
2054 M4 Pro (16c) 24 GB Qwen3.5-9B 4bit 4k 364.4 47.7 26-03-09
2055 M4 Pro (16c) 24 GB Qwen3.5-9B 4bit 1k 363.4 49.4 26-03-09
2056 M3 Ultra (80c) 512 GB Qwen3.5-9B 4bit 32k 1,315 50.1 26-03-09
2057 M4 Max (40c) 64 GB Qwen3.5-27B 4bit 1k 249.4 28.8 26-03-09
2058 M3 Ultra (80c) 512 GB Qwen3.5-9B 4bit 64k 1,139 29.4 26-03-09
2059 M4 Max (40c) 64 GB Qwen3.5-27B 4bit 4k 244.9 27.1 26-03-09
2060 M3 Ultra (80c) 512 GB Qwen3.5-9B 4bit 8k 1,465 96.2 26-03-09
3,171 results · Page 206 of 318