In a surprising announcement for the technology community, OpenAI announced that it will withdraw GPT-4.5 (the largest model of OpenAI’s digital architecture) from the API starting July 14, 2025. This decision marks a turning point in the strategy of the company, recognized for leading the development of artificial intelligence (AI).
Although GPT-4.5 will remain available to paying users through the chatbot, ChatGPT, its removal from the API raises several questions about the reasons for this change and the alternatives that will be presented to developers.
Let’s explore what GPT-4.5 is, the motivations behind its removal from the API, the differences between the models available in the API and ChatGPT, and the implications this move has for both developers and end users.
What is GPT-4.5 and what are the reasons for its withdrawal?
GPT-4.5, internally called “Orion”, was presented by OpenAI in late February 2025 as its most advanced model to date.
Designed to offer significant improvements in writing, persuasion and creativity tasks, GPT-4.5 was intended to be a powerful tool for communicative and creative applications.
While it overcame some limitations of its predecessor, GPT-4o, the model failed to position itself as the “frontier” option in industry benchmarks, standards that measure performance against current technological limits.
Why is it being removed from the API?
On the same day of the announcement, OpenAI launched GPT-4.1, a model that offers performance comparable or even superior to GPT-4.5 in key areas.
However, GPT-4.1 operates at significantly lower costs, making it a much more attractive option for developers looking to optimize their resources and reduce operating expenses.
In the API, GPT-4.5 has a price of $75 per million input tokens and $150 per million output tokens, figures that are not competitive compared to other models such as GPT-4o or GPT-4.1.
This cost difference becomes crucial when developers have to integrate models into large-scale projects, where every token counts.
The central reason is clear: optimize economic and technical efficiency. By removing GPT-4.5 from the API, OpenAI encourages developers to migrate to solutions that offer excellent performance at a much more reasonable cost.
The current offer in the API and ChatGPT
Models available in the API
The OpenAI API is the tool through which developers can integrate artificial intelligence into their applications and services. Until July 14, 2025, the following models are offered:
- GPT-4o: This is a multimodal model recognized for its speed and efficiency. Costing $2.50 per million input tokens and $10.00 per million output tokens, it is accessible for a wide variety of uses.
- GPT-4.1: Launched on April 14, 2025, this model was specifically designed to compete in terms of performance and, at the same time, be more economical. Its price is $2.00 per million tokens in and $8.00 per million tokens out.
- GPT-4.5: Although it will be available in the API until July 14, 2025, its high cost ($75.00 per million input tokens and $150.00 per million output tokens) has led OpenAI to dispense with it in this environment.
Models available on ChatGPT
ChatGPT is the user interface created by OpenAI that allows end users to interact with AI models. The offer in ChatGPT differs significantly from the API, as it adapts to the needs and expectations of the end user:
- GPT-4o: Available in all plans, from free (with certain limitations) to paid plans, such as Plus, Pro, Team and Enterprise plans.
- GPT-4.5: Maintains its exclusive presence in the ChatGPT environment, but only for paid plan subscribers. In the case of the Plus plan, it is in a research preview phase, indicating that its capabilities continue to be evaluated and fine-tuned.
An important difference is that GPT-4.1, despite being the preferred option in the API for its balance between performance and cost, is not available in ChatGPT.
This is because the ChatGPT environment focuses on delivering a rich experience for the end user, while the API is geared towards scalable and efficient solutions for developers.
The future of Artificial Intelligence at OpenAI
The recent move to remove GPT-4.5 from the API is not only an operational or economic issue, but also a sign of OpenAI’s future priorities:
- Optimization and efficiency: OpenAI is redirecting its efforts towards models that offer better performance per dollar invested. This type of decision allows the company to continue innovating without compromising the viability of its services.
- Rumors and expectations: Although no official details have been revealed yet, it is speculated that OpenAI is already working on the next generation of models, possibly under the name GPT-5. The rapid succession of launches (GPT-4o, GPT-4.5 and GPT-4.1) indicates a dynamic environment where evolution is constant.
- Sustainable innovation: The segmentation between the offer for developers and for end users suggests that OpenAI is committed to balancing technical innovation with economic sustainability, which is crucial in a field where computational costs grow exponentially.
With a more affordable API model, more companies, start-ups and independent developers will be able to integrate high-performance AI into their products, which in turn will drive competitiveness and innovation in various sectors.
This post is also available in: