OpenAI has introduced a new generation of its language models - GPT-4.1, which reported on its website, writes UNN.
Details
OpenAI has introduced its new GPT-4.1 artificial intelligence model, as well as smaller versions of GPT-4.1 mini and GPT-4.1 nano, Fast Company writes, claiming significant improvements in coding, instruction following, and long context understanding.
According to the ChatGPT manufacturer, the new models, available only in the OpenAI application programming interface (API), outperform the company's most advanced model, GPT-4o, in all respects.
With improved contextual understanding, they can support up to 1 million "tokens" - a term referring to units of data processed by an AI model. The models are also equipped with updated knowledge up to June 2024.
GPT-4.1 showed a 21% improvement compared to GPT-4o and a 27% improvement compared to GPT-4.5 in coding. Meanwhile, improvements in instruction following and long context understanding also make GPT-4.1 models more effective for supporting AI agents.
"The tests are strong, but we've focused on real utility, and developers seem very happy," CEO Sam Altman wrote in his post on social media platform X.
The company also said that the new models are significantly cheaper than GPT-4.5 and will completely replace it in the API by July 2025.