Artificial Analysis says fastest GLM-5.3 endpoints fall in rankings on total response time

Artificial Analysis reports that the GLM-5.3 endpoints with the fastest streaming take 18 to 26 seconds to return the first token, while most other endpoints take 1 to 3 seconds.

Once that wait counts toward total request time, these endpoints drop in the speed rankings.

The ranking change depends on how speed is measured. Streaming speed alone puts these endpoints at the top, but the 18 to 26 second first-token delay moves them down when total request time is counted.