🧪 GPU:

1. faster-whisper = ~1940 GB
2. tts = ~620 GB
3. Qwen/Qwen3-0.6B-FP8 = ~1360 GB
4. Qwen.Qwen3-0.6B = ~1760 GB
5. Qwen/Qwen3-1.7B = ~4240 GB
6. Qwen/Qwen2.5-0.5B = ~1200 GB
7. Total GPU = ~4450 GB (faster-whisper+tts+qwen3-0.6B)
8. Total GPU = ~6800 GB (faster-whisper+tts+qwen3-1.7B)
9. Total GPU = ~4200 GB (faster-whisper+tts+Qwen2.5-0.5B)
10. Total GPU = ~4266 GB (faster-whisper+tts+Qwen3-0.6B-FP8)
11. Total GPU = ~4254 GB (faster-whisper-medium+tts+Qwen3-0.6B-FP8)

🧪 Qwen/Qwen3-1.7B/0.6B:

1. prompt: 32K
2. Trained on same data sets (sizes)