whisper-large-v3-turbo
by openai
809M params · automatic-speech-recognition · 3.1k likes · 7.8M downloads
whisper-large-v3-turbo is a 809M parameter model. At Q4 quantization it requires 0GB of VRAM. It runs comfortably on GeForce RTX 4090 (1623 tok/s), GeForce RTX 5090 (2434 tok/s), M4 Max 128GB (594 tok/s).
Inference providers
| Provider | $/1M in | $/1M out | Throughput |
|---|---|---|---|
| HF Inference |
GPU compatibility
| GPU | VRAM | Q4 Decode | Verdict |
|---|---|---|---|
| GeForce RTX 4090 | 24GB | 1623 tok/s | comfortable |
| GeForce RTX 5090 | 32GB | 2434 tok/s | comfortable |
| M4 Max 128GB | 128GB | 594 tok/s | comfortable |
| M4 Pro 48GB | 48GB | 297 tok/s | comfortable |
| M4 Pro 24GB | 24GB | 297 tok/s | comfortable |
| A100 PCIe 80 GB | 80GB | 2973 tok/s | comfortable |
| H100 SXM5 80 GB | 80GB | 5981 tok/s | comfortable |
| GeForce RTX 3090 | 24GB | 1435 tok/s | comfortable |
| Radeon RX 7900 XTX | 24GB | 1186 tok/s | comfortable |
| GeForce RTX 4080 | 16GB | 1152 tok/s | comfortable |