distilroberta-base
by distilbert
83M params · fill-mask · 174 likes · 1.6M downloads
distilroberta-base is a 83M parameter model. At Q4 quantization it requires 0GB of VRAM. It runs comfortably on GeForce RTX 4090 (15865 tok/s), GeForce RTX 5090 (23791 tok/s), M4 Max 128GB (5805 tok/s).
Inference providers
| Provider | $/1M in | $/1M out | Throughput |
|---|---|---|---|
| HF Inference |
GPU compatibility
| GPU | VRAM | Q4 Decode | Verdict |
|---|---|---|---|
| GeForce RTX 4090 | 24GB | 15865 tok/s | comfortable |
| GeForce RTX 5090 | 32GB | 23791 tok/s | comfortable |
| M4 Max 128GB | 128GB | 5805 tok/s | comfortable |
| M4 Pro 48GB | 48GB | 2902 tok/s | comfortable |
| M4 Pro 24GB | 24GB | 2902 tok/s | comfortable |
| A100 PCIe 80 GB | 80GB | 29066 tok/s | comfortable |
| H100 SXM5 80 GB | 80GB | 58462 tok/s | comfortable |
| GeForce RTX 3090 | 24GB | 14027 tok/s | comfortable |
| Radeon RX 7900 XTX | 24GB | 11599 tok/s | comfortable |
| GeForce RTX 4080 | 16GB | 11259 tok/s | comfortable |