I Like ChatGPT 5.2 Deep Thinking
ChatGPT 5.2 with Deep Thinking is a good model (it was released yesterday)
From a coding perspective, I’ve only used GPT 5.2 for several hours today for some enhancements we are building into CEORater.
GPT 5.2 seems to be as good as Gemini 3.0 Pro and Claude Opus 4.5, but it is too early to know for sure. GPT scores higher than both Opus.4.5 and Gemini 3.0 Pro Deep Think on the benchmarks, but the benchmarks do not reflect practical usage in my experience.
Aside from the CEORater coding, I fed GPT 5.2 Deep Think the below PDF and asked it to build a spreadsheet, which it did in approximately two minutes (picture below), correctly and with nice formatting.
For comparison, Gemini dumped the numbers into Google Sheets correctly, but without any formatting. Opus 4.5 struggled for 3-4 minutes to produce anything, so I killed the exercise.
Spreadsheet creation and Presentation building were two key skills that GPT 5.2 was purposely trained for.