EuroLLM-22B-Instruct-2512
by utter-project
22.6B params · text-generation · 59 likes · 3.8k downloads
EuroLLM-22B-Instruct-2512 is a 22.6B parameter model. At Q4 quantization it requires 11GB of VRAM. It runs comfortably on GeForce RTX 4090 (58 tok/s), GeForce RTX 5090 (86 tok/s), A100 PCIe 80 GB (106 tok/s).
Inference providers
| Provider | $/1M in | $/1M out | Throughput |
|---|---|---|---|
| Public AI |
GPU compatibility
| GPU | VRAM | Q4 Decode | Verdict |
|---|---|---|---|
| GeForce RTX 4090 | 24GB | 58 tok/s | comfortable |
| GeForce RTX 5090 | 32GB | 86 tok/s | comfortable |
| M4 Max 128GB | 128GB | 21 tok/s | tight |
| M4 Pro 48GB | 48GB | 10 tok/s | tight |
| M4 Pro 24GB | 24GB | 10 tok/s | tight |
| A100 PCIe 80 GB | 80GB | 106 tok/s | comfortable |
| H100 SXM5 80 GB | 80GB | 213 tok/s | comfortable |
| GeForce RTX 3090 | 24GB | 51 tok/s | comfortable |
| Radeon RX 7900 XTX | 24GB | 42 tok/s | comfortable |
| GeForce RTX 4080 | 16GB | 41 tok/s | comfortable |