t5-small
by google-t5
61M params · translation · 531 likes · 1.8M downloads
t5-small is a 61M parameter model. At Q4 quantization it requires 0GB of VRAM. It runs comfortably on GeForce RTX 4090 (21700 tok/s), GeForce RTX 5090 (32541 tok/s), M4 Max 128GB (7940 tok/s).
Inference providers
| Provider | $/1M in | $/1M out | Throughput |
|---|---|---|---|
| HF Inference |
GPU compatibility
| GPU | VRAM | Q4 Decode | Verdict |
|---|---|---|---|
| GeForce RTX 4090 | 24GB | 21700 tok/s | comfortable |
| GeForce RTX 5090 | 32GB | 32541 tok/s | comfortable |
| M4 Max 128GB | 128GB | 7940 tok/s | comfortable |
| M4 Pro 48GB | 48GB | 3970 tok/s | comfortable |
| M4 Pro 24GB | 24GB | 3970 tok/s | comfortable |
| A100 PCIe 80 GB | 80GB | 39757 tok/s | comfortable |
| H100 SXM5 80 GB | 80GB | 79964 tok/s | comfortable |
| GeForce RTX 3090 | 24GB | 19186 tok/s | comfortable |
| Radeon RX 7900 XTX | 24GB | 15865 tok/s | comfortable |
| GeForce RTX 4080 | 16GB | 15400 tok/s | comfortable |