distilbert-base-cased-distilled-squad
by distilbert
65M params · question-answering · 266 likes · 222.0k downloads
distilbert-base-cased-distilled-squad is a 65M parameter model. At Q4 quantization it requires 0GB of VRAM. It runs comfortably on GeForce RTX 4090 (20140 tok/s), GeForce RTX 5090 (30202 tok/s), M4 Max 128GB (7370 tok/s).
Inference providers
| Provider | $/1M in | $/1M out | Throughput |
|---|---|---|---|
| HF Inference |
GPU compatibility
| GPU | VRAM | Q4 Decode | Verdict |
|---|---|---|---|
| GeForce RTX 4090 | 24GB | 20140 tok/s | comfortable |
| GeForce RTX 5090 | 32GB | 30202 tok/s | comfortable |
| M4 Max 128GB | 128GB | 7370 tok/s | comfortable |
| M4 Pro 48GB | 48GB | 3685 tok/s | comfortable |
| M4 Pro 24GB | 24GB | 3685 tok/s | comfortable |
| A100 PCIe 80 GB | 80GB | 36899 tok/s | comfortable |
| H100 SXM5 80 GB | 80GB | 74217 tok/s | comfortable |
| GeForce RTX 3090 | 24GB | 17807 tok/s | comfortable |
| Radeon RX 7900 XTX | 24GB | 14725 tok/s | comfortable |
| GeForce RTX 4080 | 16GB | 14293 tok/s | comfortable |