IDK what happened today but I used GLM-5.3 as usual from Ollama cloud and it was so fast it generated entire documents like instantly.

The reasoning and the result document were done after less than 1 or 2 seconds.

Have Ollama suddenly bought GPU capacity?