OpenAI to discontinue GPT-4.5 API, pushing developers toward GPT-4.1

OpenAI announced on Monday that it will soon wind down the availability of its largest AI model, GPT-4.5, via its API. Released in late February, GPT-4.5 will remain accessible to developers through the API until July 14, after which users will need to transition to another model in OpenAI’s catalogue. The company has identified GPT-4.1, which was launched on the same day, as the preferred replacement.

According to reports, GPT-4.1 offers similar or improved performance to GPT-4.5 in key areas, such as writing and persuasiveness, but at a significantly lower cost. The company is phasing out GPT-4.5 in order to prioritise the development of future models.

It is important to note that GPT-4.5 will continue to be available on ChatGPT, where it remains in a research preview for paying customers. The deprecation only applies to the API version.

GPT-4.5, codenamed Orion, was trained using more computing power and data than any of OpenAI’s previous models. While it offers advancements over its predecessor, GPT-4o, particularly in writing and persuasion, it does not reach 'frontier level' performance on a number of industry benchmarks.

Despite its advancements, GPT-4.5 has proven to be costly to run. OpenAI had previously warned in February that it was evaluating whether to continue offering GPT-4.5 via its API, given its high operational expenses. The pricing reflects this, with GPT-4.5 costing $75 per million input tokens (approximately 750,000 words) and $150 per million output tokens, making it one of the company's most expensive offerings.

News