On March 5, artificial intelligence (AI) company OpenAI announced the release of its GPT-5.4 Thinking model, which has begun rolling out to ChatGPT. According to OpenAI, this version has more advanced logical thinking capabilities, better knowledge to solve complex tasks, and the ability to reproduce inaccurate answers. For this reason, the updated model is designed less for everyday conversations and more for complex work, research, and document analysis.
Designed for more complex thinking
According to an OpenAI announcement, the GPT-5.4 Thinking model is designed to perform more complex multi-step thinking. This means the system is better able to break down a problem into multiple steps and provide more consistent answers.
Such skills are especially important when solving complex tasks – for example, analyzing data, developing software code, or processing large amounts of text information.
Key improvements
According to OpenAI, the new model features several important technological improvements. One of them is more advanced logical and analytical thinking, allowing it to solve complex tasks more accurately and provide more coherent explanations.
OpenAI also notes that GPT-5.4 is the company’s first major model with built-in computing capabilities. This allows AI agents to interact directly with programs, perform tasks, verify results, and automatically correct errors when necessary.
The model’s ability to handle programming tasks has also been improved. The system can generate code more accurately, analyze programming problems, and help detect potential errors in more complex projects.
Additionally, the model has improved its handling of large text and information processing. It can more effectively analyze long documents, summarize information, and extract key insights.
OpenAI also notes that the number of AI “hallucinations” – situations where the system provides inaccurate or fictional information – has been reduced. This should increase the reliability of responses when using the model on real-world work tasks.
OpenAI says that GPT-5.4 Thinking is initially available to paid ChatGPT plans and developers via the API. However, the company recently updated the free version of ChatGPT to include a faster model, GPT-5.3 Instant, which has a more natural conversational tone, shorter, clearer responses, and fewer unnecessary rejections.
ChatGPT is becoming an increasingly versatile tool
The introduction of the new model reflects OpenAI’s broader strategy to transform ChatGPT into a versatile digital assistant across a range of domains. The company aims to enable AI to go beyond answering questions and help with more complex professional tasks.
Ankstesnė šio modelio versija buvo atnaujinta dar gruodį, kai „OpenAI“ skubiai reagavo į didėjančią konkurenciją DI rinkoje, ypač iš „Google“ kuriamų „Gemini“ modelių.
Research conducted by OpenAI shows that a large proportion of users use ChatGPT for personal, not work purposes. This shows a gap between the vision of technology developers and users’ actual habits. Although OpenAI aims to make ChatGPT an increasingly universal work tool for professionals, a significant proportion of people still use it more often for everyday tasks.
Credits:
Image:
