OpenAI upgrades ChatGPT with GPT-5.4 Thinking and Pro models: What's new?

After releasing the GPT 5.3 Instant just a few days back, the AI giant has rolled out the new upgrade – GPT-5.4. OpenAI says that it is specifically to handle professional and complex tasks easily. The standard version comes along with the GPT-5.4 Thinking and a more advanced GPT-5.4 Pro variant. Also Read: ChatGPT Health missed emergency care in over half of cases: Study

The new model is now available across ChatGPT and OpenAI’s API, bringing improvements in reasoning, coding, and automation capabilities. Here is everything new that OpenAI has to offer with GPT-5.4.  Also Read: OpenAI signs Pentagon deal as Donald Trump blacklists Anthropic: Here's what happened

OpenAI GPT-5.4 with multiple versions

The tech giant has introduced the GPT-5.4 series with multiple versions depending on its usecase. It comes with a standard GPT-5.4 model, which focuses on general productivity tasks, while the GPT-5.4 Thinking is more about deeper reasoning and complex problem solving. The GPT-5.4 Pro version is for those who need higher performance and advanced capabilities. Also Read: “Work came with a price”: OpenAI senior engineer leaves over mental health

Who can try it?

At the moment, GPT-5.4 Thinking is available to ChatGPT Plus, Team, and Pro users, replacing the earlier GPT-5.2 Thinking model. However, the Pro version is limited to Pro and Enterprise subscribers for now. 

GPT-5.4: All the new upgrades

One of the big upgrades with GPT-5.4 is how much information it can handle at once. OpenAI says that in the API version, the model now supports context windows of up to 1 million tokens. In simple terms, it can read and work with much longer documents or conversations without losing track of the context. OpenAI also says the new model is more efficient with tokens. That means it can complete the same tasks using fewer tokens than earlier versions.

The company also claims GPT-5.4 performs better in several benchmark tests, especially those that simulate professional tasks such as analysing financial data, writing reports, or doing legal-style reasoning.

Add Techlusive as a Preferred Source

Another interesting addition is something OpenAI calls computer-use capability. This allows AI agents to interact with virtual computer environments. According to the company, the model can write code to control a computer, use keyboard and mouse commands, browse websites, and complete multi-step tasks. Alongside these improvements, OpenAI says it has also worked on reducing mistakes and hallucinations.