gte-tiny
by TaylorAI
23M params · sentence-similarity · 139 likes · 7.4k downloads
gte-tiny is a 23M parameter model. At Q4 quantization it requires 0GB of VRAM. It runs comfortably on GeForce RTX 4090 (57807 tok/s), GeForce RTX 5090 (86689 tok/s), M4 Max 128GB (21154 tok/s).
Inference providers
| Provider | $/1M in | $/1M out | Throughput |
|---|---|---|---|
| HF Inference |
GPU compatibility
| GPU | VRAM | Q4 Decode | Verdict |
|---|---|---|---|
| GeForce RTX 4090 | 24GB | 57807 tok/s | comfortable |
| GeForce RTX 5090 | 32GB | 86689 tok/s | comfortable |
| M4 Max 128GB | 128GB | 21154 tok/s | comfortable |
| M4 Pro 48GB | 48GB | 10577 tok/s | comfortable |
| M4 Pro 24GB | 24GB | 10577 tok/s | comfortable |
| A100 PCIe 80 GB | 80GB | 105911 tok/s | comfortable |
| H100 SXM5 80 GB | 80GB | 213021 tok/s | comfortable |
| GeForce RTX 3090 | 24GB | 51110 tok/s | comfortable |
| Radeon RX 7900 XTX | 24GB | 42266 tok/s | comfortable |
| GeForce RTX 4080 | 16GB | 41026 tok/s | comfortable |