RMBG-2.0
by briaai
221M params · image-segmentation · 1.1k likes · 242.9k downloads
RMBG-2.0 is a 221M parameter model. At Q4 quantization it requires 0GB of VRAM. It runs comfortably on GeForce RTX 4090 (5949 tok/s), GeForce RTX 5090 (8921 tok/s), M4 Max 128GB (2177 tok/s).
Inference providers
| Provider | $/1M in | $/1M out | Throughput |
|---|---|---|---|
| fal |
GPU compatibility
| GPU | VRAM | Q4 Decode | Verdict |
|---|---|---|---|
| GeForce RTX 4090 | 24GB | 5949 tok/s | comfortable |
| GeForce RTX 5090 | 32GB | 8921 tok/s | comfortable |
| M4 Max 128GB | 128GB | 2177 tok/s | comfortable |
| M4 Pro 48GB | 48GB | 1088 tok/s | comfortable |
| M4 Pro 24GB | 24GB | 1088 tok/s | comfortable |
| A100 PCIe 80 GB | 80GB | 10899 tok/s | comfortable |
| H100 SXM5 80 GB | 80GB | 21922 tok/s | comfortable |
| GeForce RTX 3090 | 24GB | 5260 tok/s | comfortable |
| Radeon RX 7900 XTX | 24GB | 4349 tok/s | comfortable |
| GeForce RTX 4080 | 16GB | 4222 tok/s | comfortable |