valid sonnet
I took a (test) subscription for glm 4.7 at z.ai. And so far the ai is ok, but I find it quite slow. Using the api directly in cline as a test is much faster. So I was wondering, is there something I can do in OpenClaw to make it faster?
I am running OpenClaw on a VM with 6 cores and 8 GB of memory running on a fast SSD. And I do not see this being stressed.
cursive zodiacBOT
rare dew
I can Share the Same Problem
valid patio
Using glm4.5air works, but glm4.7 has a default thinking mode, which I can't seem to find a way to disable.
valid sonnet
4.5 is quite a big downgrade from glm 4.7. Wish it was able to automatically choose how thinking will be, how deep.
I also have that it sometimes finished tasks but not reports anything, keeps the threat active and things you ask go to queue, while it had finished, is there a way to see if you run into token or prompt depletion?
valid patio
I also have that it sometimes finished tasks but not reports anything, keeps the...
https://github.com/serenez/openclaw/blob/main/docs/SMART-ROUTER-GUIDE-EN.md You can try my solution, which uses intelligent routing to select the most suitable model. I submitted a PR, but it was closed by the bot, which said that merging is not possible in the short term.🤣
valid sonnet
https://github.com/serenez/openclaw/blob/main/docs/SMART-ROUTER-GUIDE-EN.md You...
Thanks I will look into that.
Anybody has experiance with GLM 4.7 and MiniMax M2.1, especially how smart are they in coding and how fast are they compared?
valid patio
I only have GLM 4.7, and honestly, its performance under OpenClaw is poor, mainly due to its slow output. Even with Think mode disabled, it's still very slow. I haven't tried MiniMax yet; it's too expensive right now. 😅
coarse niche
Same challenge here.. really slow 🙁