deberta-v3-base-prompt-injection
by protectai
184M params · text-classification · 101 likes · 45.7k downloads
deberta-v3-base-prompt-injection is a 184M parameter model. At Q4 quantization it requires 0GB of VRAM. It runs comfortably on GeForce RTX 4090 (7119 tok/s), GeForce RTX 5090 (10676 tok/s), M4 Max 128GB (2605 tok/s).
Inference providers
| Provider | $/1M in | $/1M out | Throughput |
|---|---|---|---|
| HF Inference |
GPU compatibility
| GPU | VRAM | Q4 Decode | Verdict |
|---|---|---|---|
| GeForce RTX 4090 | 24GB | 7119 tok/s | comfortable |
| GeForce RTX 5090 | 32GB | 10676 tok/s | comfortable |
| M4 Max 128GB | 128GB | 2605 tok/s | comfortable |
| M4 Pro 48GB | 48GB | 1302 tok/s | comfortable |
| M4 Pro 24GB | 24GB | 1302 tok/s | comfortable |
| A100 PCIe 80 GB | 80GB | 13043 tok/s | comfortable |
| H100 SXM5 80 GB | 80GB | 26235 tok/s | comfortable |
| GeForce RTX 3090 | 24GB | 6294 tok/s | comfortable |
| Radeon RX 7900 XTX | 24GB | 5205 tok/s | comfortable |
| GeForce RTX 4080 | 16GB | 5052 tok/s | comfortable |