OpenAI launched in the middle of last month GPT-4.1 seriesAt the time, it was originally said to be API-only, so ChatGPT users couldn’t use it, but unexpectedly, just a month later, OpenAI changed course without giving any particular reason. Now all ChatGPT users, including those on the free tier, can access the latest GPT-4.1 series, which replaces the previously unlimited GPT-4o mini. This means the answers you get from now on will be somewhat better than before.
OpenAI announces that ChatGPT now offers GPT-4.1 and GPT-4.1 mini.
Earlier, OpenAI announced on its official social platforms that it is officially bringing GPT-4.1 and GPT-4.1 mini into ChatGPT. ChatGPT Plus, Pro, and Team users can now select GPT-4.1 and GPT-4.1 mini from the model picker, while Enterprise and Edu users will gain access in the coming weeks. As for the free tier of ChatGPT, GPT-4.1 mini will replace the original GPT-4o mini, which means that in the future, ChatGPT will use GPT-4.1 mini for responses, with quality superior to GPT-4o mini.
Plus, Pro, & Team users will be able to access GPT-4.1 via the "more models" dropdown in the model picker. Enterprise & Edu users will get access in the coming weeks.
We’re also introducing GPT-4.1 mini, replacing GPT-4o mini, in ChatGPT for all users.https://t.co/Hiievyk7BT
— OpenAI (@OpenAI) May 14, 2025
For those who don’t know, the free version of ChatGPT uses the GPT-4o model to reply by default, but there’s a usage limit. If you exceed it, it used to switch to the GPT-4o mini model, which was unlimited. This time, after being replaced by GPT-4.1 mini, it becomes GPT-4.1 mini with unlimited usage.
As for the difference between the two, refer to the figure below. MMLU shows the model’s multilingual intelligence; the higher the score, the smarter the model. Latency below indicates the degree of delay. Although GPT-mini’s latency is slightly higher than GPT-4o mini’s, its intelligence is much higher, almost on par with GPT-4o. In other words, the free version of ChatGPT can now almost get reply quality comparable to GPT-4o in unlimited quantities:
However, when used for programming-related tasks, the SWE-bench Verified score shows that GPT-4.1 mini still has a slight gap compared to GPT-4o (2024-11-20), but it significantly leads the previous GPT-4o mini:
From this chart, it’s also clear that for paid users, if you have programming-related needs, GPT-4.1 performs better than GPT-o3 mini (high) and GPT-4.5. GPT-4.5 is mainly strong in creativity and writing.
Another point worth noting is that the GPT 4.1 series has an extremely large context window, reaching 1,047,576 tokens, which is even more than GPT-4o’s 128,000 and o3’s 200,000, so it’s really well-suited for programming.
For the paid version, open the model menu in the top-left corner, and you’ll see the GPT-4.1 series under More Models. For the free version, there’s no need to switch—it uses GPT-4.1 mini by default.
Source: KOCPC Chinese