I noted there is a new toggle between "agent/chat" when you switch to GLM-5 on chat.z.ai. I assume this model is focused on agentic tasks. Can't wait for them to release the paper or blogpost on this. I've been very impressed with GLM-4.7, so I have high expectations on this.
The youtube channel AiCodeKing mentioned something interesting on a video where he covered this model [1]. He said GLM-5 is the first open-weight "system architect" model akin to Claude opus. He also claims it beats opus 4.6 for long running tasks. But I can't verify that. We'll see it on the Artificalanalysis benchmark.
The youtube channel AiCodeKing mentioned something interesting on a video where he covered this model [1]. He said GLM-5 is the first open-weight "system architect" model akin to Claude opus. He also claims it beats opus 4.6 for long running tasks. But I can't verify that. We'll see it on the Artificalanalysis benchmark.
1. https://www.youtube.com/watch?v=63Ail-fOrkM