OpenAI has recently launched its latest AI models, GPT-4.1, GPT-4.1 mini, and GPT-4.1 nano, which the company claims to offer improvements in coding, instruction following, and long-context comprehension. These models are currently only available through OpenAI's API (Application Programming Interface). 

According to OpenAI, GPT-4.1 processes up to 1 million tokens—units of data that determine how much information the model can analyse at once. The company states the models also incorporate updated knowledge up to June 2024. Benchmark tests indicate GPT-4.1 performs 21% better than GPT-4o and 27% better than GPT-4.5 in coding tasks.  

CEO Sam Altman noted in a post on X that while benchmarks show improvements, the focus was on real-world usability. OpenAI also highlighted cost efficiency, stating the new models operate at a "much lower cost" than GPT-4.5. The company plans to discontinue the GPT-4.5 API preview in July, citing comparable or superior performance from the GPT-4.1 series.  

The release follows OpenAI's February rollout of GPT-4.5, which was initially offered as a research preview before wider access. The latest models are positioned as more capable for AI agent applications, though independent verification of their performance claims has yet to be conducted, states a report by Reuters.