Is GPT-6 Sol just completely skill-lazy for anyone else?
retained text
I’ve just started testing GPT-6 Sol, and honestly, it’s driving me nuts. My existing skills work flawlessly with GPT-5.6. With Sol6, the very first simple task ignored the clearly matching skill and pulled data from random sources instead. After I explicitly told it to always check and fully use the appropriate skill, it immediately failed again on a basic weather request—even though the correct default location was configured inside the skill. So now I’m stuck babysitting it, checking every answer, and repeatedly asking whether it actually used the damn skill. That completely defeats the purpose. Same configuration, same skills, same commands: GPT-5.6 works, GPT-6 Sol doesn’t. Is anyone else seeing Sol6 ignore or only half-use skills like this?
document fetched 2026-09-24 00:30 · canonical https://reddit.com/r/OpenAI/comments/1wnyb67/is_gpt6_sol_just_completely_skilllazy_for_anyone