roberta-large-mnli
by FacebookAI
356M params · text-classification · 209 likes · 360.5k downloads
roberta-large-mnli is a 356M parameter model. At Q4 quantization it requires 0GB of VRAM. It runs comfortably on GeForce RTX 4090 (3683 tok/s), GeForce RTX 5090 (5524 tok/s), M4 Max 128GB (1348 tok/s).
Inference providers
| Provider | $/1M in | $/1M out | Throughput |
|---|---|---|---|
| HF Inference |
GPU compatibility
| GPU | VRAM | Q4 Decode | Verdict |
|---|---|---|---|
| GeForce RTX 4090 | 24GB | 3683 tok/s | comfortable |
| GeForce RTX 5090 | 32GB | 5524 tok/s | comfortable |
| M4 Max 128GB | 128GB | 1348 tok/s | comfortable |
| M4 Pro 48GB | 48GB | 674 tok/s | comfortable |
| M4 Pro 24GB | 24GB | 674 tok/s | comfortable |
| A100 PCIe 80 GB | 80GB | 6749 tok/s | comfortable |
| H100 SXM5 80 GB | 80GB | 13575 tok/s | comfortable |
| GeForce RTX 3090 | 24GB | 3257 tok/s | comfortable |
| Radeon RX 7900 XTX | 24GB | 2693 tok/s | comfortable |
| GeForce RTX 4080 | 16GB | 2614 tok/s | comfortable |