M5 Max 128GB vs RTX 3090
For running local LLMs · 19 reports across 5 models
Tokens per second by model
| Model | M5 Max 128GB | RTX 3090 |
|---|---|---|
| Qwen3.6up to 35B | 5.5n=2 | 104.0n=10 |
| Gemma 4up to 31B | 7.5n=3 | — |
| Qwen2.5 | — | 70.0n=2 |
| DeepSeek V3 | — | —n=1 |
| Qwen3.5up to 122B | —n=1 | — |