granite-embedding-311m-multilingual-r2
by ibm-granite
312M params · feature-extraction · 103 likes · 458.2k downloads
granite-embedding-311m-multilingual-r2 is a 312M parameter model. At Q4 quantization it requires 0GB of VRAM. It runs comfortably on GeForce RTX 4090 (4212 tok/s), GeForce RTX 5090 (6317 tok/s), M4 Max 128GB (1541 tok/s).
Inference providers
| Provider | $/1M in | $/1M out | Throughput |
|---|---|---|---|
| HF Inference |
GPU compatibility
| GPU | VRAM | Q4 Decode | Verdict |
|---|---|---|---|
| GeForce RTX 4090 | 24GB | 4212 tok/s | comfortable |
| GeForce RTX 5090 | 32GB | 6317 tok/s | comfortable |
| M4 Max 128GB | 128GB | 1541 tok/s | comfortable |
| M4 Pro 48GB | 48GB | 770 tok/s | comfortable |
| M4 Pro 24GB | 24GB | 770 tok/s | comfortable |
| A100 PCIe 80 GB | 80GB | 7718 tok/s | comfortable |
| H100 SXM5 80 GB | 80GB | 15524 tok/s | comfortable |
| GeForce RTX 3090 | 24GB | 3724 tok/s | comfortable |
| Radeon RX 7900 XTX | 24GB | 3080 tok/s | comfortable |
| GeForce RTX 4080 | 16GB | 2989 tok/s | comfortable |