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 ↓ |
|---|---|---|---|---|---|---|---|---|
| 11 | M3 Ultra (60c) | 256 GB | Ornith-1.0-35B | bf16 | 128k | 839.1 | 44.4 | 26-07-21 |
| 12 | M3 Ultra (60c) | 256 GB | Ornith-1.0-35B | bf16 | 195k | 615.2 | 37.5 | 26-07-21 |
| 13 | M3 Ultra (60c) | 256 GB | Ornith-1.0-35B | bf16 | 64k | 1,267 | 54.0 | 26-07-21 |
| 14 | M3 Ultra (60c) | 256 GB | Ornith-1.0-35B | bf16 | 32k | 1,699 | 61.0 | 26-07-21 |
| 15 | M4 Max (40c) | 128 GB | Astrea-R8-Chat-9B | unknown | 4k | 965.0 | 28.5 | 26-07-21 |
| 16 | M3 Ultra (60c) | 256 GB | Ornith-1.0-35B | bf16 | 16k | 2,005 | 65.6 | 26-07-21 |
| 17 | M4 Max (40c) | 128 GB | Astrea-R8-Chat-9B | unknown | 1k | 956.6 | 28.7 | 26-07-21 |
| 18 | M5 Pro (20c) | 64 GB | Qwen3.6-27B | 6bit | 64k | 303.5 | 10.0 | 26-07-21 |
| 19 | M4 Pro (16c) | 24 GB | Qwen2.5-Coder-7B-Instruct | 4bit | 16k | 346.2 | 41.6 | 26-07-21 |
| 20 | M4 Pro (16c) | 24 GB | Qwen2.5-Coder-7B-Instruct | 4bit | 4k | 410.1 | 51.5 | 26-07-21 |