Upgrades to Codex — gpt-5-codex

i’ve just checked and i can confirm i have model gpt-5-codex available to my API account

BOOO

Error: This model is only supported in v1/responses and not in v1/chat/completions.

When will websearch be added to the ide extension? i can set he flag in the toml config but it never uses the websearch tool.

I just tried GPT-5-Codex in VS code and it’s FAST! Wow, it just flies through reasoning.

My only feedback is that it doesn’t summarize all the things it did, like GPT-5 does.

Since yesterday, I have been experiencing an issue with Codex Nube: upon completing the patching process, it erases all the code in the file and only inserts the patch into the original file. Previously, it would search for the error within the entire code and replace only the relevant section. Now, however, it deletes everything and leaves only the patch. This happens when you look for the new PR on GitHub.

Translated by AI.
Original post

Desde ayer estoy experimentando un problema con codex nube y es que al finalizar el parche del archivo, borra todo el código del archivo y solo coloca el parche en el archivo original. Antes buscaba el fallo en el código completo y reemplazaba esa parte ahora borra todo y deja el parche eso ocurre cuando buscas el nuevo pr en github

Update on September 23, 2025: GPT‑5-Codex is now available to developers using Codex via API key (in addition to being available to developers using Codex via their ChatGPT subscription).

https://openai.com/index/introducing-upgrades-to-codex/

1 Like

Feedback on the new Codex version (Web)

I don’t know what is happening with Codex, but this new version really has serious problems.
Before it was released, my workflow speed and output quality were 5–10x better. Even the previous version was more accurate and consistent.

Now I have to double-check even the task descriptions Codex is generating, because too often they come out wrong. On top of that, I don’t need Codex to spend 1–2–5 hours “working” on tasks when the final results are still broken or incomplete.

I’m using the web version, and compared to the earlier one, this feels like a clear downgrade in both speed and reliability.