DeBERTa-v3-large-mnli-fever-anli-ling-wanli
by MoritzLaurer
435M params · zero-shot-classification · 119 likes · 177.9k downloads
DeBERTa-v3-large-mnli-fever-anli-ling-wanli is a 435M parameter model. At Q4 quantization it requires 0GB of VRAM. It runs comfortably on GeForce RTX 4090 (3017 tok/s), GeForce RTX 5090 (4525 tok/s), M4 Max 128GB (1104 tok/s).
Inference providers
| Provider | $/1M in | $/1M out | Throughput |
|---|---|---|---|
| HF Inference |
GPU compatibility
| GPU | VRAM | Q4 Decode | Verdict |
|---|---|---|---|
| GeForce RTX 4090 | 24GB | 3017 tok/s | comfortable |
| GeForce RTX 5090 | 32GB | 4525 tok/s | comfortable |
| M4 Max 128GB | 128GB | 1104 tok/s | comfortable |
| M4 Pro 48GB | 48GB | 552 tok/s | comfortable |
| M4 Pro 24GB | 24GB | 552 tok/s | comfortable |
| A100 PCIe 80 GB | 80GB | 5529 tok/s | comfortable |
| H100 SXM5 80 GB | 80GB | 11121 tok/s | comfortable |
| GeForce RTX 3090 | 24GB | 2668 tok/s | comfortable |
| Radeon RX 7900 XTX | 24GB | 2206 tok/s | comfortable |
| GeForce RTX 4080 | 16GB | 2141 tok/s | comfortable |