OpenAI has unveiled a new family of language understanding LLM models called GPT-4.1 for developers to use through the API. The family includes GPT-4.1, GPT-4.1 mini, and GPT-4.1 nano models. All three can process up to 1 million tokens in context, surpassing the previous limit of 128 thousand tokens set by GPT-4o.
"We have trained GPT-4.1 to reliably perceive information throughout a 1 million token long context. We also trained it to be much more reliable than GPT-4o in noticing relevant text and ignoring distracting factors across both long and short contexts," the company OpenAI announced.
The GPT-4.1 models feature enhanced coding skills and instruction-following abilities, making fewer errors compared to their predecessors. The cost of using these new models is 26% lower than that of GPT-4o. All models in the GPT-4.1 family are available exclusively to developers via API; they are not accessible through ChatGPT.