Medical-NER
by blaze999
184M params · token-classification · 225 likes · 14.8k downloads
Medical-NER is a 184M parameter model. At Q4 quantization it requires 0GB of VRAM. It runs comfortably on GeForce RTX 4090 (7139 tok/s), GeForce RTX 5090 (10707 tok/s), M4 Max 128GB (2612 tok/s).
Inference providers
| Provider | $/1M in | $/1M out | Throughput |
|---|---|---|---|
| HF Inference |
GPU compatibility
| GPU | VRAM | Q4 Decode | Verdict |
|---|---|---|---|
| GeForce RTX 4090 | 24GB | 7139 tok/s | comfortable |
| GeForce RTX 5090 | 32GB | 10707 tok/s | comfortable |
| M4 Max 128GB | 128GB | 2612 tok/s | comfortable |
| M4 Pro 48GB | 48GB | 1306 tok/s | comfortable |
| M4 Pro 24GB | 24GB | 1306 tok/s | comfortable |
| A100 PCIe 80 GB | 80GB | 13081 tok/s | comfortable |
| H100 SXM5 80 GB | 80GB | 26310 tok/s | comfortable |
| GeForce RTX 3090 | 24GB | 6312 tok/s | comfortable |
| Radeon RX 7900 XTX | 24GB | 5220 tok/s | comfortable |
| GeForce RTX 4080 | 16GB | 5067 tok/s | comfortable |