Skip to content

Pick your style: Denkbaar

Book a conversation

Articles

History

GPT-4.1 focuses on code and long input

·

OpenAI introduced GPT-4.1 in the API with support for up to one million tokens of context.

An employee finds relevant passages in a long manual.
A large context does not remove the need to choose relevant material. Illustration: Denkbaar.

On 14 April 2025, OpenAI introduced GPT-4.1, GPT-4.1 mini and GPT-4.1 nano in the API. The models supported up to one million tokens of context. Tokens are pieces of text that a model processes. OpenAI emphasised programming and following instructions. The announcement describes the family. A long manual lies before a wide input tray and three model modules.

For a builder, a million tokens immediately raises a practical question: what material does the application actually need? A large context window provides room to supply a lot of material. It does not prove that a specific question will be answered correctly. Organising the material and its assessment remains a separate design choice. A full manual lies beside three selected passages and an hourglass.

As a trial, a small or medium-sized business could ask questions about a long public manual. Choose questions whose answers are spread throughout the document. Request the supporting passage with each answer. Also check a question for which the manual contains no information. Alongside quality, measure waiting time and the amount of text sent. Compare this approach with a trial that supplies only relevant passages. Let the results determine how much context your application actually needs.

Sources

  1. GPT-4.1 (1M context, API)

Topics: model