OpenAI opened ChatGPT API

OpenAI has announced that it now allows third party developers to integrate ChatGPT into their apps and services via an API. In addition, the company has opened up Whisper, its AI-powered speech-to-text model, and made it available through an API as well.

OpenAI says the ChatGPT API can be used for more than just creating AI-based chats (although it highlights several companies that have used it for this purpose, including the Snap My AI feature announced earlier this week). The developers say the new model family, called gpt-3.5-turbo, is “the best model for many non-chat use cases.”

OpenAI is offering 1,000 tokens for $0.002 and says it’s “10x cheaper than current GPT-3.5 models” due in part to “a series of system-wide optimizations.” While 1000 is a lot, it’s worth noting that sending a single piece of text for an API response can cost several tokens (“tokens” are blocks of text into which the system breaks sentences and words to predict what text it should output).

According to the OpenAI documentation, for example, the request “ChatGPT is great!” takes six tokens – the API breaks it down into “Chat”, “G”, “PT”, “is”, “great” and “!”. The company provides a tool to check how many tokens it will take to interpret a line of text, and says a general rule of thumb is that “one token usually equates to about 4 characters” in English.

The company says that developers will also be able to get a dedicated instance of ChatGPT if they process a large amount of data through the API. The post says this will give you more control over which model you use, how quickly you want the AI ​​to respond to requests, and how long conversations with the bot can last.

OpenAI also announced another new API for Whisper, a speech-to-text model. The company says you can use it to translate audio to text for $0.006 per minute. Whisper is technically open source, so you can run it on your own hardware without paying anything. However, OpenAI likely has access to more powerful hardware, so if you’re looking for speed or need to transcribe on low power devices like phones, using the API might be the way to go.

The company also says it’s working on improving uptime and that “its engineering team’s top priority now is the stability of production use cases.”

While a few developers have come up with workarounds to include chat services in their apps, including using OpenAI’s regular GPT API that has been available for a while, the introduction of the official ChatGPT API could lead to a new wave of AI apps. While there are many companies working on their own AI chatbot models, such things are completely out of reach for most developers. Now they can simply use the OpenAI technology.