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 ↓
263401 M3 Max (30c) 36 GB Qwen3.5-9B 4bit 1k 580.5 53.9 26-03-21
263402 M3 Ultra (60c) 256 GB Qwen3.5-9B 4bit 1k 1,125 101.7 26-03-21
263403 M4 Pro (20c) 64 GB Qwen3.5-9B-Claude-4.6-Opu... 4bit 1k 401.4 48.4 26-03-21
263404 M4 Pro (20c) 64 GB Qwen3.5-9B-Claude-4.6-Opu... 4bit 4k 434.3 46.6 26-03-21
263405 M4 Pro (20c) 64 GB Qwen3.5-9B-Claude-4.6-Opu... 4bit 8k 433.0 44.7 26-03-21
263406 M3 Max (30c) 36 GB Llama-3.2-3B-Instruct unknown 4k 1,342 38.3 26-03-21
263407 M3 Max (30c) 36 GB Llama-3.2-3B-Instruct unknown 1k 1,258 40.2 26-03-21
263408 M4 (10c) 16 GB Qwen3.5-9B 4bit 1k 221.9 19.7 26-03-21
263409 M4 (10c) 16 GB Qwen3.5-9B 4bit 4k 223.0 18.9 26-03-21
263410 M3 (10c) 16 GB Qwen3.5-9B 4bit 4k 185.3 17.3 26-03-21
292,916 results · Page 26341 of 29292