granite-embedding-97m-multilingual-r2
by ibm-granite
97M params · feature-extraction · 119 likes · 291.5k downloads
granite-embedding-97m-multilingual-r2 is a 97M parameter model. At Q4 quantization it requires 0GB of VRAM. It runs comfortably on GeForce RTX 4090 (13474 tok/s), GeForce RTX 5090 (20207 tok/s), M4 Max 128GB (4930 tok/s).
Inference providers
| Provider | $/1M in | $/1M out | Throughput |
|---|---|---|---|
| HF Inference |
GPU compatibility
| GPU | VRAM | Q4 Decode | Verdict |
|---|---|---|---|
| GeForce RTX 4090 | 24GB | 13474 tok/s | comfortable |
| GeForce RTX 5090 | 32GB | 20207 tok/s | comfortable |
| M4 Max 128GB | 128GB | 4930 tok/s | comfortable |
| M4 Pro 48GB | 48GB | 2465 tok/s | comfortable |
| M4 Pro 24GB | 24GB | 2465 tok/s | comfortable |
| A100 PCIe 80 GB | 80GB | 24687 tok/s | comfortable |
| H100 SXM5 80 GB | 80GB | 49654 tok/s | comfortable |
| GeForce RTX 3090 | 24GB | 11913 tok/s | comfortable |
| Radeon RX 7900 XTX | 24GB | 9852 tok/s | comfortable |
| GeForce RTX 4080 | 16GB | 9563 tok/s | comfortable |