paraphrase-multilingual-MiniLM-L12-v2
by sentence-transformers
118M params · sentence-similarity · 1.2k likes · 23.1M downloads
paraphrase-multilingual-MiniLM-L12-v2 is a 118M parameter model. At Q4 quantization it requires 0GB of VRAM. It runs comfortably on GeForce RTX 4090 (11159 tok/s), GeForce RTX 5090 (16735 tok/s), M4 Max 128GB (4083 tok/s).
Inference providers
| Provider | $/1M in | $/1M out | Throughput |
|---|---|---|---|
| HF Inference |
GPU compatibility
| GPU | VRAM | Q4 Decode | Verdict |
|---|---|---|---|
| GeForce RTX 4090 | 24GB | 11159 tok/s | comfortable |
| GeForce RTX 5090 | 32GB | 16735 tok/s | comfortable |
| M4 Max 128GB | 128GB | 4083 tok/s | comfortable |
| M4 Pro 48GB | 48GB | 2041 tok/s | comfortable |
| M4 Pro 24GB | 24GB | 2041 tok/s | comfortable |
| A100 PCIe 80 GB | 80GB | 20446 tok/s | comfortable |
| H100 SXM5 80 GB | 80GB | 41123 tok/s | comfortable |
| GeForce RTX 3090 | 24GB | 9866 tok/s | comfortable |
| Radeon RX 7900 XTX | 24GB | 8159 tok/s | comfortable |
| GeForce RTX 4080 | 16GB | 7920 tok/s | comfortable |