bart-large-mnli
by facebook
407M params · zero-shot-classification · 1.5k likes · 3.8M downloads
bart-large-mnli is a 407M parameter model. At Q4 quantization it requires 0GB of VRAM. It runs comfortably on GeForce RTX 4090 (3223 tok/s), GeForce RTX 5090 (4833 tok/s), M4 Max 128GB (1179 tok/s).
Inference providers
| Provider | $/1M in | $/1M out | Throughput |
|---|---|---|---|
| HF Inference |
GPU compatibility
| GPU | VRAM | Q4 Decode | Verdict |
|---|---|---|---|
| GeForce RTX 4090 | 24GB | 3223 tok/s | comfortable |
| GeForce RTX 5090 | 32GB | 4833 tok/s | comfortable |
| M4 Max 128GB | 128GB | 1179 tok/s | comfortable |
| M4 Pro 48GB | 48GB | 589 tok/s | comfortable |
| M4 Pro 24GB | 24GB | 589 tok/s | comfortable |
| A100 PCIe 80 GB | 80GB | 5905 tok/s | comfortable |
| H100 SXM5 80 GB | 80GB | 11877 tok/s | comfortable |
| GeForce RTX 3090 | 24GB | 2849 tok/s | comfortable |
| Radeon RX 7900 XTX | 24GB | 2356 tok/s | comfortable |
| GeForce RTX 4080 | 16GB | 2287 tok/s | comfortable |