GPT 5.2 - Out Now!

Discussions  Featured Discussions
1

OpenAI announcement · Pricing

OpenAI just dropped GPT 5.2! It’s available in Cursor now.

Any surprises?

2 Likes
GPT 5.2 is now available in Cursor!
2

Good first prompt responses :wink:

3

I’m confused… why is it so fast (for GPT)…? …like Claude-fast in that first text coming through as a response that feels way more natural than 5.1. I don’t want to get my hopes up…

gpt52-hello

2 Likes
4

image
image913×550 46.3 KB

curious about opus 4.5 real world task comparisons

5 Likes
6

what!? is gpt 5.2 that much better than gemini 3 pro and opus 4.5!? have u tried in coding too?

7

What about you? How’s everything going?

8

i haven’t tried gpt 5.2 models yet ( afk

9

where is cursor discount?

4 Likes
10

Just did one test, it directly solved some issues opus 4.5 that i love didn’t manage in 4 tries. Will try more tomorrow!

2 Likes
11

It seems amazing in debug mode.

1 Like
12

excited to dig in here!

13

Is there a free usage period? :sweat_smile:

11 Likes
14

You should give 1 week free usage for every new model

4 Likes
15

It keeps freezing when I open the terminal. Is this the same for everyone?

16

顶顶顶顶顶

17

I’m quite tired of the constant model release and naming. How does it compare to GPT-5.1-Codex Max High? Anyone compared?

18

After testing a little more it seems good, im testing extra high fast. But its way slower then opus 4.5

19

It works very slowly.

20

Nah not that good, I’ll stick with Opus :innocent:

21

I was holding back a comment like this because it isn’t well-informed enough in regards to GPT-5.2 experience, but it seems like an ok share based on other thread comments:

  • GPT-5.2 coding-model probably coming early next year, so no comparison with GPT-5.1 Codex, …is what I hear. I always had better results with non-Codex models anyway
  • Gemini was terrible for me. I turned it off after a couple of days of trying it and always undoing its work (real coding here; in a large codebase)
  • Opus 4.5 is my daily driver right now, especially via Claude Code VSCode extension being basically unlimited usage at the $100 tier so far. I’ve even used its plans to drive Cursor agent implementations with Composer and then review. …this a pattern I want to try with GPT-5.2…maybe as a reviewer.
  • GPT-5.2 seems really good though so far, but isn’t as conversational (it doesn’t feel like a team-member sitting with me pair-programming). I want to use it more in Debug-mode. So far I am loving Debug.