bert-base-chinese
by google-bert
103M params · fill-mask · 1.4k likes · 603.6k downloads
bert-base-chinese is a 103M parameter model. At Q4 quantization it requires 0GB of VRAM. It runs comfortably on GeForce RTX 4090 (12762 tok/s), GeForce RTX 5090 (19138 tok/s), M4 Max 128GB (4670 tok/s).
Inference providers
| Provider | $/1M in | $/1M out | Throughput |
|---|---|---|---|
| HF Inference |
GPU compatibility
| GPU | VRAM | Q4 Decode | Verdict |
|---|---|---|---|
| GeForce RTX 4090 | 24GB | 12762 tok/s | comfortable |
| GeForce RTX 5090 | 32GB | 19138 tok/s | comfortable |
| M4 Max 128GB | 128GB | 4670 tok/s | comfortable |
| M4 Pro 48GB | 48GB | 2335 tok/s | comfortable |
| M4 Pro 24GB | 24GB | 2335 tok/s | comfortable |
| A100 PCIe 80 GB | 80GB | 23382 tok/s | comfortable |
| H100 SXM5 80 GB | 80GB | 47028 tok/s | comfortable |
| GeForce RTX 3090 | 24GB | 11283 tok/s | comfortable |
| Radeon RX 7900 XTX | 24GB | 9331 tok/s | comfortable |
| GeForce RTX 4080 | 16GB | 9057 tok/s | comfortable |