同一款恐怖游戏分别用 Opus 5.5、Sonnet 5.5 和 GPT 6.1 Sol 做对比

Nic Vandewetering @nvandewetering原推发布于 2026.10.01

在 X 上打开原推

okay so GPT 6.1 Sol edited this entire livestream highlights video. Pretty great. but the games that came out of it are disappointing. I have built separate Saw horror game on Claude Opus 5.5, Sonnet 5.5, and now ChatGPT 6.1 Sol. Sol is by far the worst outcome. overall I’d say it’s an okay model, but given the politics around Dev Day and reduced usage limits, I really don’t have interest in using this model beyond the usage credits and reset that I was given. what results have you guys gotten with ChatGPT 6.1 Sol?