OpenAI introduced a new GPT-5.4 model for ChatGPT, API, and Codex

OpenAI has announced the launch of a new artificial intelligence model, GPT-5.4. It has become the most powerful and efficient advanced model for professional work and is available in ChatGPT, API, and Codex. This is reported by UNN with reference to OpenAI.

Details

Today we are releasing GPT‑5.4 in ChatGPT (as GPT‑5.4 Thinking), API, and Codex. This is our most powerful and efficient advanced model for professional work. We are also releasing GPT‑5.4 Pro in ChatGPT and API for those who want maximum performance on complex tasks.

According to the company, GPT-5.4 combines the latest advances in reasoning, programming, and agentic workflows into a single model. It leverages the coding capabilities of GPT-5.3-Codex and improves performance across various tools and software environments.

The model can work with spreadsheets, presentations, and documents. According to the developers, this allows for more accurate and efficient execution of complex professional tasks.

In ChatGPT, the new GPT-5.4 Thinking version can show a preliminary plan of its actions. This allows users to adjust the direction of the response while the model is working and get a result that better meets their needs.

The model also improves deep web research, especially for highly specialized queries, and better maintains context in complex issues that require longer analysis.

In Codex and API, GPT-5.4 has gained the ability to use a computer, allowing agents to control computers and perform complex workflows across various applications. The model supports up to 1 million tokens of context, enabling it to plan, execute, and verify long-term tasks.

Recall

OpenAI recalled the GPT-4o update due to the chatbot's excessive friendliness. The company promises to fix this as soon as possible to avoid awkward and stressful situations for users.