Advisor / Can it run?
Can my hardware run this model?
Pairs where we have measured data, with speed, memory fit and context limits. Only combinations with their own measurements get a page.
20 measured pairs
| Hardware | Model | Gen tok/s | Fit | Origin |
|---|---|---|---|---|
| RTX 4090 24GB | gpt-oss 20B | 258.4 | fits fully | own run |
| RTX 4090 24GB | Qwen3.6 35B-A3B | 190.5 | fits fully | own run |
| RTX 4090 24GB | Qwen3-Coder 30B-A3B | 266 | fits fully | own run |
| RTX 4090 24GB | Gemma 4 31B | 44.4 | fits fully | own run |
| RTX 4090 24GB | Qwen3.6 27B | 48.4 | fits fully | own run |
| RTX 4090 24GB | Gemma 4 26B-A4B | 169.3 | fits fully | own run |
| RTX 3090 24GB | gpt-oss 20B | 204.2 | fits fully | own run |
| RTX 3090 24GB | Qwen3.6 35B-A3B | 159 | fits fully | own run |
| RTX 3090 24GB | Qwen3-Coder 30B-A3B | 206.3 | fits fully | own run |
| RTX 3090 24GB | Gemma 4 31B | 38.8 | fits fully | own run |
| RTX 3090 24GB | Qwen3.6 27B | 42.2 | fits fully | own run |
| RTX 3090 24GB | Gemma 4 26B-A4B | 140.3 | fits fully | own run |
| RTX 5090 32GB | Qwen3.6 27B | 80.7 | fits fully | public runs |
| AMD Ryzen AI Max+ 395 (Strix Halo) 128GB | gpt-oss 120B | 56.4 | fits fully | public runs |
| NVIDIA DGX Spark 128GB | gpt-oss 120B | 56.8 | fits fully | public runs |
| 16GB VRAM class (RTX 4060 Ti 16GB, 32GB RAM) | gpt-oss 20B | 94.4 | fits fully | own run |
| RTX 5070 12GB | Gemma 4 12B | 70.1 | fits fully | own run |
| RTX 4070 12GB | Qwen3.5 9B | 51.3 | fits fully | own run |
| RTX 3060 12GB (32GB RAM) | Qwen3.5 9B | 36.6 | fits fully | own run |
| RTX 3090 24GB | Gemma 4 12B | 57.2 | fits fully | own run |