bert-base-german-cased
by google-bert
110M params · fill-mask · 82 likes · 482.4k downloads
bert-base-german-cased is a 110M parameter model. At Q4 quantization it requires 0GB of VRAM. It runs comfortably on GeForce RTX 4090 (11968 tok/s), GeForce RTX 5090 (17948 tok/s), M4 Max 128GB (4379 tok/s).
Inference providers
| Provider | $/1M in | $/1M out | Throughput |
|---|---|---|---|
| HF Inference |
GPU compatibility
| GPU | VRAM | Q4 Decode | Verdict |
|---|---|---|---|
| GeForce RTX 4090 | 24GB | 11968 tok/s | comfortable |
| GeForce RTX 5090 | 32GB | 17948 tok/s | comfortable |
| M4 Max 128GB | 128GB | 4379 tok/s | comfortable |
| M4 Pro 48GB | 48GB | 2189 tok/s | comfortable |
| M4 Pro 24GB | 24GB | 2189 tok/s | comfortable |
| A100 PCIe 80 GB | 80GB | 21927 tok/s | comfortable |
| H100 SXM5 80 GB | 80GB | 44103 tok/s | comfortable |
| GeForce RTX 3090 | 24GB | 10581 tok/s | comfortable |
| Radeon RX 7900 XTX | 24GB | 8750 tok/s | comfortable |
| GeForce RTX 4080 | 16GB | 8494 tok/s | comfortable |