Openai launches three models much cheaper via the API © Pexels
OPENAI announces the launch of three new models: GPT-4.1, GPT-4.1 Mini and GPT-4.1 NANO. Although the deployment of new models is increasingly frequent in the firm of Sam Altman, these three newborns are not intended to replace GPT-4O. The latter, currently in place and acting as a default model in the chatgpt interface, should however be rebuilt in a few months. The GPT-5 model should then succeed him; It will represent the evolution of GPT-4.5-Currently on test-deemed too expensive at the request for the moment.
Three new models for the Openai API
If many publishers relayed the rumor that the GPT-4.1, GPT-4.1 mini and GPT-4.1 Nano models aimed to replace GPT-4O, the information proved to be false. These three new models are intended for developers and will thus be accessible only through the Openai API. Concretely, GPT-4.1 and its variations represent a lower cost of use, are faster and accept much more tokens.
GPT-4.1 vs GPT-4O-screenshot © OPENAI
The GPT-4.1 models support up to 1 million tokens, representing around 750,000 words. As a result, Openai offers a capacity for a contextual window similar to that of Google with its Gemini model. For comparison, GPT-4O is confined to 128,000 tokens. Likewise, GPT-4.1 is doing better than the 4O model on many benchmarks, with the exception of requests that require manipulating the image or the voice.
On the side of the prices applied by Openai on the API and calculated by million tokens, GPT-4o costs $ 2.5 input and $ 10 output. GPT-4.1 returns to $ 2 inputs for 8 dollars out. Version 4.1 Mini costs $ 0.40 at a starter and $ 1.60 output. GPT-4.1 NANO broke all records at OPENAI since the model invoices $ 0.10 per million tokens at the start and $ 0.40 output.
Counter competition with developers
In short, this model is less, but it is better and for a more affordable price. OPENAI is also transparent on this point: Sam Altman explicitly indicates that GPT-4.1 has more important needs in terms of contextualization of a prompt to function properly.
GPT -4.1 (and -mini and -nano) are now available in the API!
These models are Great at Coding, Following, and Long Context instruction (1 million tokens).
Benchmarks are strong, but we focused on real-world utility, and developers seem very happy.
GPT-4.1 Family is API-ONLY.
– Sam Altman (@sama) April 14, 2025
The models of the 4.1 series are also optimized to excel in the tasks linked to computer coding, as indicated by the CEO of Openai in person.
With these three new models, Openai sends a clear message to developers in order to encourage them to use their models rather than promoting competition which, so far, proposed more affordable solutions.
For Chatgpt end users through the OpenAi platform, the major evolution will only be done in a few months, when the firm deploys GPT-5 as a default model.
