gte-modernbert-base
by Alibaba-NLP
149M params · sentence-similarity · 192 likes · 219.5k downloads
gte-modernbert-base is a 149M parameter model. At Q4 quantization it requires 0GB of VRAM. It runs comfortably on GeForce RTX 4090 (8811 tok/s), GeForce RTX 5090 (13213 tok/s), M4 Max 128GB (3224 tok/s).
Inference providers
| Provider | $/1M in | $/1M out | Throughput |
|---|---|---|---|
| HF Inference |
GPU compatibility
| GPU | VRAM | Q4 Decode | Verdict |
|---|---|---|---|
| GeForce RTX 4090 | 24GB | 8811 tok/s | comfortable |
| GeForce RTX 5090 | 32GB | 13213 tok/s | comfortable |
| M4 Max 128GB | 128GB | 3224 tok/s | comfortable |
| M4 Pro 48GB | 48GB | 1612 tok/s | comfortable |
| M4 Pro 24GB | 24GB | 1612 tok/s | comfortable |
| A100 PCIe 80 GB | 80GB | 16143 tok/s | comfortable |
| H100 SXM5 80 GB | 80GB | 32469 tok/s | comfortable |
| GeForce RTX 3090 | 24GB | 7790 tok/s | comfortable |
| Radeon RX 7900 XTX | 24GB | 6442 tok/s | comfortable |
| GeForce RTX 4080 | 16GB | 6253 tok/s | comfortable |