3 ms·
It's cheap because it's a Flash model, far smaller and much less compute for inference, runs on TPUs instead of GPUs.
by panarky 2y ago
It's cheap because it's a Flash model, far smaller and much less compute for inference, runs on TPUs instead of GPUs.