OpenAI just launched a new set of AI models called the GPT-4.1 family, and they’re a big step up from previous versions.
These models—GPT-4.1, GPT-4.1 mini, and GPT-4.1 nano—are now available to developers through OpenAI’s API (the naming scheme could get better me thinks). They improve on everything from coding to understanding long documents, and they’re also cheaper and faster to use.
The flagship model, GPT-4.1, is built for heavy tasks like software engineering, analyzing huge documents, or building AI tools that can follow complex instructions. It can handle massive inputs—up to 1 million tokens (basically, a lot of text or code)—and performs better than earlier versions in areas like writing code, following detailed instructions, and understanding long or complicated information.
For example, it scored nearly 55% on a coding benchmark called SWE-bench, way ahead of previous models like GPT-4o, which scored around 33%. SWE-bench is a test that checks how well an AI can fix real bugs in real software projects by reading code and writing working solutions. That makes GPT-4.1 one of the best AI tools for real-world programming right now. It’s also better at creating clean code edits and building web apps with better design and functionality.
Instruction-following is another big improvement. GPT-4.1 does a better job listening to what you actually want it to do—like following formats, avoiding certain words, or keeping track of things in multi-step conversations. For tasks where you tell the AI, “Don’t do this” or “Say it in this order,” GPT-4.1 is much more accurate, saving you time from re-prompting.
OpenAI also released GPT-4.1 mini and nano, smaller versions that are much cheaper and faster. The nano version, for instance, is OpenAI’s quickest and most affordable model so far, good for simple tasks like sorting data or autocomplete.
These new models are ideal for developers building tools that need to handle large amounts of information quickly and accurately—like customer support agents, document analyzers, or tax and legal assistants.
While GPT-4.1 isn’t available directly inside ChatGPT yet, OpenAI says many of the improvements will show up in future updates to GPT-4o, which powers the ChatGPT experience for most users.
GPT-4.5, a previous version meant for testing, will be shut down in July now that GPT-4.1 is more powerful and efficient.
Interestingly enough, OpenAI Sam Altman has responded to criticism of the GPT naming model. “How about we fix our model naming by this summer and everyone gets a few more months to make fun of us (which we very much deserve) until then?” said Altman. That sounds more than fair.
Want to see more of our stories on Google?
P.S. Want to keep this site truly independent? Support us by buying us a beer, treating us to a coffee, or shopping through Amazon here. Links in this post are affiliate links, so we earn a tiny commission at no charge to you. Thanks for supporting independent Canadian media!
Open AI is actually getting dumber. Try asking it to do some basic arithmetic such as adding up 5 different sets of long numbers. Even Bard/Gemini couldn't get the right answer, but surprisingly Bing's Co-Pilot did. Similarly, ChatGPT can't get Excel formulas right on the first try. Co-Pilot did. Gemini couldn't provide an accurate formula for Google Sheets, but Co-Pilot, once again did.
I didn’t expect much when I started this online job, but last month I made $18,000 just using my laptop at home. The work is super easy—mostly copy-paste stuff—and I do it in my free time. I don’t have to talk to anyone or follow strict hours. They pay every week, and it’s always on time. If you want something simple that actually pays, check this out.
Open This Web…….. PROMAX50.VERCEL.APP
4.o was way dumber than 3 so hopefully this fixes that