Qwen3-32B-4bit
3 workload results across 2 hardware configurations.
Fastest known config
34.4 decode tok/s
on M3 Ultra (60-core GPU) + 96GB unified via mlx — see full run
M3 Pro (18-core GPU) + 36GB unified
| Workload | Backend | Quant | decode tok/s | prefill tok/s | TTFT | Run |
|---|---|---|---|---|---|---|
| chat-short | mlx@0.31.3 | — | 7.16tok/s | 25.65tok/s | 4,288ms | r_pnrrpcdqfo4 |
| chat-short | mlx@0.31.3 | — | no data | no data | no data | r_cnyw_pb3guv |
M3 Ultra (60-core GPU) + 96GB unified
| Workload | Backend | Quant | decode tok/s | prefill tok/s | TTFT | Run |
|---|---|---|---|---|---|---|
| chat-short | mlx@0.31.3 | — | 34.41tok/s | 95.12tok/s | 1,156ms | r_anmmc80-aoq |