GPT-5.2 - Frontier model for professional work and long-running agents
by•
GPT‑5.2 is designed to unlock even more economic value for people; it’s better at creating spreadsheets, building presentations, writing code, perceiving images, understanding long contexts, using tools, and handling complex, multi-step projects.
Replies
Does everyone feel how it's become more stupid each time?
Honestly, yes.
It's quite good compared to GPT-5.1 in terms of daily coding and creative writing. But I'd still use Claude for the real difficult coding tasks, and Gemini for multimodal tasks.
Does that mean that GPT-5.1 tokens are becoming cheaper?
Work with the agents has finally become productive. I was really worried about how agents handled multi-step instructions in version 5.1! Nice
I’ve tried both 5.1 — honestly, the code quality hasn’t impressed me.