all-MiniLM-L6-v2
by sentence-transformers
23M params · sentence-similarity · 4.6k likes · 202.1M downloads
all-MiniLM-L6-v2 is a 23M parameter model. At Q4 quantization it requires 0GB of VRAM. It runs comfortably on GeForce RTX 4090 (57806 tok/s), GeForce RTX 5090 (86687 tok/s), M4 Max 128GB (21153 tok/s).
Inference providers
| Provider | $/1M in | $/1M out | Throughput |
|---|---|---|---|
| HF Inference |
GPU compatibility
| GPU | VRAM | Q4 Decode | Verdict |
|---|---|---|---|
| GeForce RTX 4090 | 24GB | 57806 tok/s | comfortable |
| GeForce RTX 5090 | 32GB | 86687 tok/s | comfortable |
| M4 Max 128GB | 128GB | 21153 tok/s | comfortable |
| M4 Pro 48GB | 48GB | 10576 tok/s | comfortable |
| M4 Pro 24GB | 24GB | 10576 tok/s | comfortable |
| A100 PCIe 80 GB | 80GB | 105909 tok/s | comfortable |
| H100 SXM5 80 GB | 80GB | 213016 tok/s | comfortable |
| GeForce RTX 3090 | 24GB | 51109 tok/s | comfortable |
| Radeon RX 7900 XTX | 24GB | 42265 tok/s | comfortable |
| GeForce RTX 4080 | 16GB | 41025 tok/s | comfortable |