bge-reranker-large
by BAAI
560M params · feature-extraction · 451 likes · 704.0k downloads
bge-reranker-large is a 560M parameter model. At Q4 quantization it requires 0GB of VRAM. It runs comfortably on GeForce RTX 4090 (2345 tok/s), GeForce RTX 5090 (3516 tok/s), M4 Max 128GB (858 tok/s).
Inference providers
| Provider | $/1M in | $/1M out | Throughput |
|---|---|---|---|
| HF Inference |
GPU compatibility
| GPU | VRAM | Q4 Decode | Verdict |
|---|---|---|---|
| GeForce RTX 4090 | 24GB | 2345 tok/s | comfortable |
| GeForce RTX 5090 | 32GB | 3516 tok/s | comfortable |
| M4 Max 128GB | 128GB | 858 tok/s | comfortable |
| M4 Pro 48GB | 48GB | 429 tok/s | comfortable |
| M4 Pro 24GB | 24GB | 429 tok/s | comfortable |
| A100 PCIe 80 GB | 80GB | 4296 tok/s | comfortable |
| H100 SXM5 80 GB | 80GB | 8641 tok/s | comfortable |
| GeForce RTX 3090 | 24GB | 2073 tok/s | comfortable |
| Radeon RX 7900 XTX | 24GB | 1714 tok/s | comfortable |
| GeForce RTX 4080 | 16GB | 1664 tok/s | comfortable |