Short Shelf Life: OpenAI Sunsets Flagship GPT-4.5 API Access, Cites Cost and New 4.1 Model
▼ Summary
– OpenAI will discontinue access to its GPT-4.5 model via API by July 14th, 2025, due to cost and performance considerations.
– GPT-4.5, codenamed “Orion,” was launched in late February and is known for its high resource consumption and expensive API pricing.
– OpenAI is promoting the newer GPT-4.1 model as a cost-effective and similarly performant successor to GPT-4.5.
– The discontinuation of GPT-4.5 API access allows OpenAI to reallocate resources towards developing future models.
– GPT-4.5 will remain available within ChatGPT for paying subscribers, but developers using the API must transition to GPT-4.1 or other models.
In a move highlighting the rapid iteration cycles and potential cost pressures in large-scale AI deployment, OpenAI announced it will soon discontinue access to its largest model, GPT-4.5, via its API. The model, codenamed “Orion” during development, only launched through the API in late February, giving it a remarkably short tenure as a top-tier offering for developers.
Developers relying on GPT-4.5 will have until July 14th, 2025, to transition their applications. After that date, API calls to the model will no longer function.
The ‘Why’: Cost vs. Performance
OpenAI isn’t leaving developers empty-handed. The company is explicitly positioning its newly launched GPT-4.1 model family as the preferred successor. According to an OpenAI spokesperson, the newer GPT-4.1 offers “similar or improved performance than GPT-4.5 in key areas at a much lower cost.”
This cost factor appears central to the decision. GPT-4.5 was known internally and acknowledged externally to be extremely resource-intensive to run. Its API pricing reflected this, set at a steep $75 per million input tokens and $150 per million output tokens – significantly higher than other models in OpenAI’s portfolio. Even back in February, OpenAI hinted that the long-term availability of GPT-4.5 via the API was under evaluation due to these operational expenses.
By deprecating GPT-4.5 API access, OpenAI stated it aims to “prioritize building future models,” suggesting a strategic reallocation of computational resources towards development rather than maintaining the costly infrastructure for this specific high-end model via the API.
GPT-4.5’s Brief Reign
Trained with reportedly more data and computing power than any prior OpenAI model, GPT-4.5 demonstrated improvements over its predecessor, GPT-4o, particularly in tasks involving sophisticated writing and persuasion. Despite its scale, however, reports indicated it didn’t consistently reach “frontier level” performance across all standard industry benchmarks compared to some competitors or theoretical expectations.
It’s crucial to note that GPT-4.5 isn’t disappearing entirely. OpenAI confirmed the model will remain available within ChatGPT for paying subscribers participating in its research preview. The discontinuation is strictly limited to the API access used by developers building their own applications on top of OpenAI’s technology.
Implications for Developers
The swift retirement of a flagship API model underscores the dynamic nature of the AI platform landscape. Developers who integrated GPT-4.5 for its specific capabilities, perhaps leveraging its strengths in complex generation tasks despite the cost, now face a migration deadline. While GPT-4.1 is presented as a capable replacement, developers will need to test and potentially adapt their applications to ensure compatibility and desired performance levels with the newer model.
This move signals OpenAI’s ongoing efforts to balance cutting-edge performance with practical operational efficiency and cost-effectiveness in its API offerings, steering developers towards models that offer a more sustainable combination of power and price.
(Source: TechCrunch)
