Openai manuscript upgrade with a new version of GPT-5

Openai announced on Monday that it releases a new version of GPT-5 to artificial intelligence coding agent, Manuscript. The company says its new model, called GPT-5-Codex, spends its dynamic “thinking” more than previous models, and can spend anywhere in a few seconds to seven hours on the coding mission. As a result, it works better on the coding standards.

The new model is now launched in Codex products – which can be accessed via an IDE, GitHub or ChatGPT products – for all Chatgpt Plus, PRO, Business, EDU and Enterprise users. Openai says it is planning to make the model to provide the API customers in the future.

The update is part of Openai’s efforts to make Codex more competitive with other artificial intelligence coding products, such as Claude Code, Anysphere’s Cursor or Microsoft Github Copilot. The artificial intelligence coding tools market has become more crowded last year, as a result of an intense user request. Indicator He exceeded $ 500 million in Arr Earlier in 2025 and Windsurf, a similar symbol editor, it was a subject Attempt to seize chaotic That witnessed his team’s division between Google and perception.

Openai says that GPT-5-Codex outperforms GPT-5 On The bench has been checkedCoding coding capabilities, as well as standard performance performance in the tasks of reshaping code from large warehouses, in force.

The company also says it has trained GPT-5-Codex to conduct software review reviews, and has requested an experience of software engineers to assess the review review comments. According to what was reported, engineers found GPT-5-Codex to provide fewer incorrect comments, with more “high-influential comments”.

In the briefing, Alexander Egrikos, by Alexander Egrikos at Openai, said that a lot of increased performance was thanks to the “thinking capabilities” in GPT-5-Codex. Users may be familiar with the guidance of the GPT-5 in ChatGPT, which directs information to various models based on the complexity of the task. Embiricos said that the GPT-5-Codex works similarly, but it has no router under the cap, and it can adjust the time that work in the actual time.

This is an advantage compared to the router, which decides the amount of calculations and time to use it in a problem. Instead, GPT-5-Codex can decide five minutes in a problem that needs to spend another hour. Embiricos said he saw that the model took seven hours in some cases.