OpenAI, the company behind ChatGPT, is aware that its artificial intelligence model remains the benchmark in a market increasingly saturated with alternatives. With strong competitors like Copilot and Gemini gaining ground, the company remains vigilant and continues to invest in constant innovation. Proof of this is the imminent launch of its new models: GPT-4.1, GPT-4.1 mini, and GPT-4.1 nano.
These new versions are now available through the company’s API and will soon be available to all users through the official app.
New ChatGPT models: what they look like
Although not yet integrated into ChatGPT, these models represent such a significant advance that there’s already talk of a paradigm shift, especially in the field of programming and the handling of long contexts. And with good reason: the new GPT-4.1 significantly surpasses its predecessors, including the acclaimed GPT-40 and GPT-4.5 itself, which is officially obsolete with this update.
GPT-4.1’s greatest strength lies in its ability to understand and generate code with unprecedented accuracy. In tests, this new model showed a 21.4% improvement over GPT-4.1 and a 26.6% improvement over GPT-4.5 in programming tasks. This not only means better results, but also a significant leap in quality when it comes to following complex instructions and understanding extended contexts, essential in software development environments or large-scale data analysis.

One of the most striking points is its ability to handle up to a million context tokens, an astronomical figure compared to what previous models could handle. This improvement allows GPT-4.1 to understand and process large documents, entire codebases, or long conversations without losing the thread. To put it simply, the more context tokens a model can handle, the more “memory” it has to reason and work with large amounts of information without losing the thread.
More tokens, more intelligence
Furthermore, OpenAI has not limited itself to presenting a single version of this model. The company has also launched GPT-4.1 mini and GPT-4.1 nano, two more limited versions that maintain much of the power of the main model, but with advantages in terms of cost and speed. GPT-4.1 mini matches, and even exceeds in some cases, the performance of GPT-40, but with much lower latency and an 83% lower cost. For its part, GPT-4.1 nano is presented as the fastest and most economical solution, perfect for tasks such as text classification, auto-completion, or simple pattern detection.

The prices also reflect OpenAI’s ambition to democratize access to these models. GPT-4.1 costs $2 per million input tokens and $8 per million output tokens. The mini version reduces these costs to $0.40 and $1.60, respectively, while the nano version costs just $0.10 per million input tokens and $0.40 per million output tokens.
It’s important to note that while GPT-4.1 is not yet available to ChatGPT users, many of its improvements are already being incorporated into GPT-4.0, and this integration is expected to expand in the coming months. All models share the same cutoff date: June 2024.






