bert-base-multilingual-uncased-sentiment
by nlptown
167M params · text-classification · 465 likes · 1.3M downloads
bert-base-multilingual-uncased-sentiment is a 167M parameter model. At Q4 quantization it requires 0GB of VRAM. It runs comfortably on GeForce RTX 4090 (7845 tok/s), GeForce RTX 5090 (11765 tok/s), M4 Max 128GB (2870 tok/s).
Inference providers
| Provider | $/1M in | $/1M out | Throughput |
|---|---|---|---|
| HF Inference |
GPU compatibility
| GPU | VRAM | Q4 Decode | Verdict |
|---|---|---|---|
| GeForce RTX 4090 | 24GB | 7845 tok/s | comfortable |
| GeForce RTX 5090 | 32GB | 11765 tok/s | comfortable |
| M4 Max 128GB | 128GB | 2870 tok/s | comfortable |
| M4 Pro 48GB | 48GB | 1435 tok/s | comfortable |
| M4 Pro 24GB | 24GB | 1435 tok/s | comfortable |
| A100 PCIe 80 GB | 80GB | 14373 tok/s | comfortable |
| H100 SXM5 80 GB | 80GB | 28910 tok/s | comfortable |
| GeForce RTX 3090 | 24GB | 6936 tok/s | comfortable |
| Radeon RX 7900 XTX | 24GB | 5736 tok/s | comfortable |
| GeForce RTX 4080 | 16GB | 5567 tok/s | comfortable |