distilbert-base-uncased-distilled-squad
by distilbert
66M params · question-answering · 119 likes · 56.8k downloads
distilbert-base-uncased-distilled-squad is a 66M parameter model. At Q4 quantization it requires 0GB of VRAM. It runs comfortably on GeForce RTX 4090 (19784 tok/s), GeForce RTX 5090 (29669 tok/s), M4 Max 128GB (7240 tok/s).
Inference providers
| Provider | $/1M in | $/1M out | Throughput |
|---|---|---|---|
| HF Inference |
GPU compatibility
| GPU | VRAM | Q4 Decode | Verdict |
|---|---|---|---|
| GeForce RTX 4090 | 24GB | 19784 tok/s | comfortable |
| GeForce RTX 5090 | 32GB | 29669 tok/s | comfortable |
| M4 Max 128GB | 128GB | 7240 tok/s | comfortable |
| M4 Pro 48GB | 48GB | 3620 tok/s | comfortable |
| M4 Pro 24GB | 24GB | 3620 tok/s | comfortable |
| A100 PCIe 80 GB | 80GB | 36248 tok/s | comfortable |
| H100 SXM5 80 GB | 80GB | 72906 tok/s | comfortable |
| GeForce RTX 3090 | 24GB | 17492 tok/s | comfortable |
| Radeon RX 7900 XTX | 24GB | 14465 tok/s | comfortable |
| GeForce RTX 4080 | 16GB | 14041 tok/s | comfortable |