OpenAI launches GPT-4.1 AI model with big coding-focused upgrades

Just a heads up, if you buy something through our links, we may get a small share of the sale. It’s one of the ways we keep the lights on here. Click here for more.

OpenAI has just released a new artificial intelligence model called GPT-4.1, which is an upgrade to its previous version, GPT-4o. 

According to OpenAI, GPT-4.1 is especially better at writing code and following instructions, making it more useful for developers and general users alike. 

One of the standout improvements is its ability to handle a much larger amount of information at once, up to one million tokens. That’s a huge jump from the previous limit of 128,000 tokens in GPT-4o. 

OpenAI says they trained GPT-4.1 to stay focused on relevant information even when given very long pieces of text or complex inputs like images or videos. 

Along with the main GPT-4.1 model, OpenAI also released two smaller versions: GPT-4.1 Mini and GPT-4.1 Nano. 

These are lighter, faster, and cheaper models for developers who may not need the full power of GPT-4.1. 

The Nano version is the smallest and most affordable AI model OpenAI has made so far. 

In fact, GPT-4.1 is 26% cheaper to use than GPT-4o, which is important in a competitive AI market where other companies are releasing very efficient models. 

OpenAI is planning to retire its older GPT-4 model from ChatGPT by April 30th, and it will also phase out the GPT-4.5 preview by July 14th. 

The company believes GPT-4.1 performs just as well, or even better, than these older models, but at lower cost and faster speeds. 

GPT-4o, which recently got image-generation features, became so popular that OpenAI had to pause free user access due to high demand. 

Lastly, CEO Sam Altman confirmed that GPT-5 will be delayed by a few months because combining all the new features into one model has been more difficult than expected. 

Meanwhile, OpenAI is also preparing to launch new “reasoning” models called o3 and o4 mini soon.

Do you use ChatGPT for coding? Do you see yourself using this model? We want to hear your thoughts in the comments below, or via our Twitter or Facebook.