OpenAI представила нову модель GPT-5.4: що відомо
OpenAI has unveiled its new GPT-5.4 language model.
In addition to the basic version, users also have access to GPT-5.4 Pro and GPT-5.4 Thinking.
The Pro version is optimised for high performance. At the same time, the Thinking version is designed to perform more complex tasks that require reasoning.
The API version of the new model has a context window of up to 1 million tokens.
The company notes that this is the highest figure among OpenAI language models.
According to the developers, GPT-5.4 can perform the same tasks using significantly fewer tokens than the previous GPT-5.2 model.
The new version also features an updated system for working with tools in the API.
In particular, the Tool Search mechanism allows the model to find tool definitions only when they are needed, instead of passing all instructions at once.
The company also reported improvements in the accuracy of responses.
According to their data, GPT-5.4 makes 33% fewer errors in individual statements and contains 18% fewer inaccuracies overall compared to the previous model.
