I had a very productive useful day, but was working on ~8 things at once.
I guess I do prefer slow but better than fast but ok. GLM-4.6 was pretty alright, but I had a hard time trusting it, felt like it took a lot of my time in any trickier area.
Now yes the time to first token is being somewhat brutal. Its hard to wait for. But damn is the output just so much better. It's so much better about really digging in to reference materials, is more aggressively seeing the lay of the land, and the output quality is so much higher.
It sucks, yes, is definitely a bit annoying especially when compared to how fast so many other services are. But I still will say, I treasure that step up in output enough that I would absolutely not go back, if this is just how it is now. On the net the trust ability I have here now with GLM-5 makes that generally a pretty easy tradeoff for me!
I should probably try adding a Flash model to my OpenCode though!! Cause sometimes I just need the thing done and it's not hard!!!!
I guess I do prefer slow but better than fast but ok. GLM-4.6 was pretty alright, but I had a hard time trusting it, felt like it took a lot of my time in any trickier area.
Now yes the time to first token is being somewhat brutal. Its hard to wait for. But damn is the output just so much better. It's so much better about really digging in to reference materials, is more aggressively seeing the lay of the land, and the output quality is so much higher.
It sucks, yes, is definitely a bit annoying especially when compared to how fast so many other services are. But I still will say, I treasure that step up in output enough that I would absolutely not go back, if this is just how it is now. On the net the trust ability I have here now with GLM-5 makes that generally a pretty easy tradeoff for me!
I should probably try adding a Flash model to my OpenCode though!! Cause sometimes I just need the thing done and it's not hard!!!!