Hacker News Reader

Running inference on one of these models takes like a GPU minute, so they can't just let the public use them.

They can absolutely do for it if they charge the public the cost of GPU time.

Can't be this; Google Colab gives out tons of free GPU usage.

Google has a lot of GPUs, but even so Colab seems like it’s a lot cheaper than it should be. You can get some very good GPUs on the paid plan.