3 ms·
I have one specific task where GPT-3.5 failed completely, but GPT-4 succeeded spectacularly (generating correctly formatted AutoRig Pro bone mappings in Blender
by nmfisher 4y ago
I have one specific task where GPT-3.5 failed completely, but GPT-4 succeeded spectacularly (generating correctly formatted AutoRig Pro bone mappings in Blender from one armature to another).
4 still fails regularly on a lot on seemingly basic tasks, but it is a noticeable step up from 3.5. As they continue to scale it up, I suggest checking back in every few months to see if the newer versions perform any better.