gemma-2-9b-it-SimPO
by princeton-nlp
9.2B params · text-generation · 172 likes · 431 downloads
gemma-2-9b-it-SimPO is a 9.2B parameter model. At Q4 quantization it requires 5GB of VRAM. It runs comfortably on GeForce RTX 4090 (142 tok/s), GeForce RTX 5090 (213 tok/s), M4 Max 128GB (51 tok/s).
Inference providers
| Provider | $/1M in | $/1M out | Throughput |
|---|---|---|---|
| Featherless |
GPU compatibility
| GPU | VRAM | Q4 Decode | Verdict |
|---|---|---|---|
| GeForce RTX 4090 | 24GB | 142 tok/s | comfortable |
| GeForce RTX 5090 | 32GB | 213 tok/s | comfortable |
| M4 Max 128GB | 128GB | 51 tok/s | comfortable |
| M4 Pro 48GB | 48GB | 25 tok/s | tight |
| M4 Pro 24GB | 24GB | 25 tok/s | tight |
| A100 PCIe 80 GB | 80GB | 260 tok/s | comfortable |
| H100 SXM5 80 GB | 80GB | 523 tok/s | comfortable |
| GeForce RTX 3090 | 24GB | 125 tok/s | comfortable |
| Radeon RX 7900 XTX | 24GB | 103 tok/s | comfortable |
| GeForce RTX 4080 | 16GB | 100 tok/s | comfortable |