Qwen3-VL-4B-Instruct
10 records across 3 hardware.
| hardware | model | quant | backend | tok/s | pp tok/s | W | ttft s | ctx/tg | date | class | flags | source |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| M5 Max | Qwen3-VL-4B-Instruct | 4bit | oMLX | 141.5 | 4739.0 | – | – | 1024 | 2026-10-15 | community | – | source |
| M5 Max | Qwen3-VL-4B-Instruct | 4bit | oMLX | 118.2 | 4218.0 | – | – | 4096 | 2026-10-15 | community | – | source |
| M1 Pro | Qwen3-VL-4B-Instruct | 4bit | oMLX | 54.1 | 360.4 | – | – | 1024 | 2026-10-11 | community | – | source |
| M1 Pro | Qwen3-VL-4B-Instruct | 4bit | oMLX | 54.1 | 360.4 | – | – | 1024 | 2026-10-11 | community | – | source |
| M1 Pro | Qwen3-VL-4B-Instruct | 4bit | oMLX | 42.4 | 328.8 | – | – | 4096 | 2026-10-11 | community | – | source |
| M1 Pro | Qwen3-VL-4B-Instruct | 4bit | oMLX | 42.4 | 328.8 | – | – | 4096 | 2026-10-11 | community | – | source |
| M4 | Qwen3-VL-4B-Instruct | 4bit | oMLX | 39.6 | 326.7 | – | – | 1024 | 2026-10-10 | community | – | source |
| M4 | Qwen3-VL-4B-Instruct | 4bit | oMLX | 39.6 | 326.7 | – | – | 1024 | 2026-10-10 | community | – | source |
| M4 | Qwen3-VL-4B-Instruct | 4bit | oMLX | 26.0 | 292.4 | – | – | 4096 | 2026-10-10 | community | – | source |
| M4 | Qwen3-VL-4B-Instruct | 4bit | oMLX | 26.0 | 292.4 | – | – | 4096 | 2026-10-10 | community | – | source |