Hi all! We’re rolling out new versions of GPT-3 and Codex ! These models introduce two new capabilities, Insert and Edit that allow you to change existing text, rather than just completing it. Insert The new insert capability adds contextually relevant text in the middle of existing content, by pr…

I have been using AI editing tools for over a year. Like StudioAI and Jasper. I would give the edit tool a 7 unless my writing is so perfect it needs almost cero re-edits. Which is not the case as english is my second language and other tools give me better choices. Some times it did surprise me wit…

The result isn’t impressive in French. The limit of token is also very low. Useful for an email, but not really for a professional document, or would need too many manipulation from the user to be time saving.

I’m experiencing the same thing. Text isn’t that long, but reducing input length dramatically appears to fix this. I suppose it makes sense since it’s free currently, but this makes it impossible to test with text of any meaningful length, imo.

I used the codex model to fix code that had multiple errors. It was quite impressive and produced mostly fixed code however it didn’t manage to do a few simple things like closing brackets.

Yeah I thought so too, was very disappointed since edit seems to be very useful, oddly enough it’s been something that normal GPT could always manage quite well, so not sure why it’s so finnicky as its own function.

Can we fine tune the edit model to improve editing with samples of pairs of unedited and edited text?

Not very impressed with the edit capabilities (at least for my 3000 token document). Ideal prompt: Reduce the size of the document, while keeping the narrative. Indicate areas to refine by wrapping them in << >>, and areas that can be removed by wrapping them in [].

It would be helpful to have a way to determine what was changed. This is specially helpful in long texts so proofing the result can be more manageable.

Hi, a few suggestions for Insert and Edits that would really help in any use case: Diff generation with red/green highlighting to see what the model changed in the text Scrolling both input/output windows at once. Scrolling each window individually is a major pain, especially for long inputs/outp…

I was just messing around with the editing capability and it generated the f-slur (f**, as a part of “furf**”) without it getting marked by the content-warning filter? When will content filter support come to the edit feature

Introducing Insert and Edits Capabilities

Announcements

Topic		Replies	Views
Big Idea: GPT as a universal concept translator Community idea	46	2001	September 5, 2024
GPT-5 Coding Feels Downgraded — Please Fix This Codex	127	16561	January 27, 2026
A theoretical novel writing tool API	29	4929	November 23, 2021
Prompt Fatigue Question For API Calls Prompting gpt-35-turbo	24	1014	January 25, 2025
It looks like GPT-4-32k is rolling out API gpt-4	202	73279	July 16, 2023

Introducing Insert and Edits Capabilities

Related topics