OpenAI brings its GPT-4 1 models to ChatGPT

Initially intended for use only by third-party software and AI developers through OpenAI’s application programming interface (API), GPT-4.1 was added to ChatGPT following strong user feedback. OpenAI says the O4 mini is best used for “fast technical tasks,” like quick STEM-related queries. It says it’s also ideal for visual reasoning, like extracting key data points from a CSV file or providing a quick summary of a scientific article.

OpenAI’s new GPT-4.1 models can process a million tokens and solve coding problems better than ever

“We will see even more models over time as cost drops, open source increases, and specialized models win out in different arenas including biology, chip design, and more,” he says. Etzioni adds that he would like to see companies focus on reducing the cost and environmental impact of training powerful models in the years ahead. All of the new models can analyze eight times more code at once, which improves their ability to make improvements and fix bugs. The new models are also better at following instructions given by users, reducing the need to repeat commands in different ways to get the desired result. OpenAI showed demos of GPT-4.1 building different apps including a flashcard app for language learning.

Datadog President Amit Agarwal on Trends in…

GPT-4 stood as one of the brand’s most popular and longest-running large language models. However, the company has already shifted its focus away from its original large language model technology and more toward its series of reasoning models and other technologies in recent months. While OpenAI has shown an affinity for reasoning models, it has also continued to develop and release its LLMs. The brand made a statement by launching GPT-4.5 in late February 2025 as its largest AI model to date.

Chiplet technology enables manufacturers to assemble multiple smaller semiconductor dies into a single package, allowing for better yields and reducing production costs. The Turing test is considered important because it allows researchers to study the potential applications and ethical ramifications of AI in real time. The study determined that through the GPT-4.5 model, humans could be able to have successful conversations with AI.

Chinese AI startup DeepSeek also recently upgraded its models, putting additional pressure on OpenAI to maintain its leadership position. On its internal OpenAI-MRCR test, accuracy dropped from around 84% with 8,000 tokens to 50% with one million tokens. Financial firm Carlyle reported 50% better performance on extracting granular financial data from dense documents — a critical capability for investment analysis and decision-making.

Making GPT-4o the default model is a step towards that effort to streamline its products. For those who have been following OpenAI and the AI industry, it puts the breakneck speed of the industry into sharp relief, while simultaneously pointing out that GPT-5 has yet to emerge. GPT-4, released in March 2023, was a notable step up from GPT-3.5, the previous model, which ushered in the ChatGPT explosive introduction to the world.

A report on the state of AI published by Stanford University this month found that models from Google and DeepSeek now have similar capabilities to models from OpenAI. It also showed a gaggle of other firms including Anthropic, Meta, and the French firm Mistral in close pursuit. On today’s livestream, Varun Mohan, CEO of Windsurf, a popular tool for AI coding, said that the company had been testing GPT-4.1 and found that the new model was “60 percent” better than GPT-4o according to its own benchmarks. “We found that GPT-4.1 has substantially fewer cases of degenerate behavior,” Mohan said, noting that the new model spends less time reading and editing irrelevant files by mistake. GPT-4.1 scored 55 percent on SWE-Bench, a widely used benchmark for gauging the prowess of coding models. The new models are “great at coding, they’re great at complex instruction following, they’re fantastic for building agents,” Weil said.

Unlike those released as part of OpenAI’s “o” series, GPT-4.5 is not considered to be a reasoning model. The startup is not just looking into reasoning models, but users can expect to see a more blended experience overall with future releases for ChatGPT where you don’t have to pick which one to use. However, it is making room for its planned model by ending support for other models in its lineup.

GPT‑4.5 will begin rolling out to Plus and Team users next week, then to Enterprise and Edu users the following week. Due to GPT-4.5’s massive alleged size, does it become even harder to parse what’s going on inside of the model? Ryder doesn’t think system interpretability, the attempt to understand why a model generates specific outputs, will be harder due to scaling. Actually, he sees the same methods used for smaller models directly applying to these more massive endeavors.

Last week, Altman announced that ChatGPT would receive a memory upgrade allowing the chatbot to better remember and refer back to previous conversations. In late March, Altman announced that OpenAI plans to release an open-weight model, which developers will be able to download and modify for free, in the summer. Open-weight models are already popular with researchers, developers, and startups because they can be tailored for different uses and are often cheaper to use. The GPT-4.1 models should help software engineers who are using ChatGPT to write or debug code, OpenAI spokesperson Shaokyi Amdo told TechCrunch. GPT-4.1 excels at coding and instruction following compared to GPT-4o, according to OpenAI, but is faster than its o-series of reasoning models.

• The model, which was released over two years ago, will still be available in the API, but recent updates to GPT-4o have rendered GPT-4 somewhat obsolete.
• The test entails human participants selecting which model they thought was human and which they thought was AI after five minutes of engagement.
• However, it is making room for its planned model by ending support for other models in its lineup.

However, OpenAI announced its decision to retire GPT-4.5 in mid-April, just months after making the model available. The company said it didn’t meet the “frontier level” of AI that it was aiming for, with technology exceeding current models. OpenAI plans to fully remove GPT-4.5 from its model lineup, having advised developers that they will have access to the API until July 14, and giving guidance to transition their projects to other APIs. “GPT‑4.1 is a significant step forward in the practical application of AI,” OpenAI said last month when introducing the newest family of AI models in the API. OpenAI faces pressure to show that it can build a sustained and profitable business by selling access to its AI models to other companies.