OpenAI has presented another major update, as the fifth update is now available since the introduction of GPT-5. 5.1 had eliminated some criticisms of GPT-5 at the end of 2025, which was followed by the rather unsuccessful 5.2 release, which was replaced by 5.3 (standard) and 5.4 (thinking) in March. The new version is called GPT-5.5, this time in both instant and thoughtful modes. As usual, the announcement was made via a blog entry that goes into detail about improvements and results in common AI benchmarks. Areas OpenAI focused on
The title as the most intelligent and intuitive model to date is a standard formulation of the company that is used with every update. However, OpenAI also names specific topic areas on which it has particularly focused. According to OpenAI 5.5 is particularly good at writing and debugging code, conducting online research, evaluating data, creating documents and tables, and operating software. The model should also better deal with unclear, multi-stage tasks and turn them into usable results.
Codex was the focus
In the blog entry, OpenAI also highlights Codex’s improvements – this is no coincidence, as this is a discipline in which there is strong competition from Claude Opus, among others. GPT-5.5 can better identify how a system is structured, where an error actually occurs, which part of the code is affected and what side effects a change could have.
Rollout and token prices
As usual, a new GPT model will be launched for the Plus, Pro, Business and Enterprise tariffs, with a short delay for all users. GPT‑5.5 and GPT‑5.5 Pro will also be available via the API “shortly”. The monthly prices of the subscriptions remain unchanged, but anyone who pays per million tokens as an API developer will be asked to pay $5 instead of $2.50 (input) or $30 instead of $15 (output) – even though GPT-5.5 requires significantly fewer tokens for comparable tasks, as OpenAI explains.
