OpenAI has launched its latest AI model, GPT-4.1, along with smaller versions GPT-4.1 mini and GPT-4.1 nano, delivering significant improvements in coding capabilities, instruction following, and long-context comprehension.
The new models are available exclusively through OpenAI’s application programming interface (API).
According to the company, GPT-4.1 outperforms the current GPT-4o model across all major metrics. In coding benchmarks, GPT-4.1 showed a 21% improvement over GPT-4o and a 27% gain over GPT-4.5. The new models are designed to handle up to 1 million tokens, enabling them to process larger and more complex inputs effectively. They also feature updated knowledge through June 2024.
“Benchmarks are strong, but we focused on real-world utility, and developers seem very happy,” OpenAI CEO Sam Altman said in a post on social media platform X.
OpenAI emphasized that the new models operate at a “much lower cost” than GPT-4.5. As a result, the company plans to discontinue the GPT-4.5 preview currently available in the API by July, citing improved or comparable performance from GPT-4.1.
GPT-4.5 was introduced in February as a research preview for select developers, with plans to expand access over time. With the release of GPT-4.1, OpenAI is shifting toward broader deployment of models optimized for both technical strength and cost efficiency.