OpenAI Upgrades API With GPT-4 and Opens DALL-E and Whisper API Access
OpenAI’s GPT-4 large language model (LLM) is now available through the generative AI startup’s API, widening access to the company’s most advanced text-generating algorithm. API customers in good standing can now employ GPT-4 and the additional power and features it provides, while older models accessible through the API will start to be phased out. OpenAI is also making the DALL-E 2 image generator model and Whisper speech-to-text model generally available through the API.
GPT-4 enhances the text and code generation abilities of its predecessor and adds new capabilities to the mix, including the ability to process images as input, though that’s currently limited to OpenAI’s partner Be My Eyes at the moment. The updated model has also proven capable of passing high-level academic tests given in medical and law schools. That said, incorrect and imaginary responses still pop up when using GPT-4, and the AI has plenty of blind spots. Still, it does bring a lot more power to projects employing the API. Not every company needs that much power (or wants to pay the additional cost).
That’s partly why OpenAI is also introducing GPT-3.5 Turbo to the API. Turbo serves as a midway point between GPT-3.5 and GPT-4 and has been the model fueling ChatGPT, Bing AI, and other marquee generative AI projects. Developers will have a chance to fine-tune both models using their own data starting later this year, which should help overcome some of the shortfalls.
“GPT-4 is our most capable model. Millions of developers have requested access to the GPT-4 API since March, and the range of innovative products leveraging GPT-4 is growing every day,” OpenAI wrote in a blog post about the plan. “We plan to continue investing most of our platform efforts in this direction, as we believe it will offer an increasingly capable and easy-to-use experience for developers. We’re working on closing the last few remaining gaps of the Chat Completions API quickly, such as log probabilities for completion tokens and increased steerability to reduce the “chattiness” of responses.”
GPT-4 is the marquee name, but it’s no small thing for OpenAI to make DALL-E 2 and Whisper APIs generally available. OpenAI said the image and speech generative AI models passed its benchmarks for “stability and readiness.” As the new models come in, OpenAI is balancing the demand on its servers by deprecating some of its older versions beginning next year.
“In the coming weeks, we will reach out to developers who have recently used these older models, and will provide more information once the new completion models are ready for early testing,” OpenAI wrote. “We recognize this is a significant change for developers using those older models. Winding down these models is not a decision we are making lightly. We will cover the financial cost of users re-embedding content with these new models. We will be in touch with impacted users over the coming days.”