OpenAI announces open API so developers can integrate ChatGPT into their products

On March 2, OpenAI announced that it now allows third-party developers to integrate ChatGPT into their applications and services through API, so that will be much cheaper than using existing language models.

Additionally, the company said it is opening up the API for Whisper, an AI-powered speech-to-text model, and making some important changes to its developer terms of service.

OpenAI says its ChatGPT API can be used for more than just creating an AI-powered chat interface, although it acknowledges that several companies have been It's being used for similar purposes, including Snap's My AI feature announced earlier this week and Microsoft's previous chat feature for Bing.

It is worth noting that this model opened by OpenAI may not be the one used by Bing, because Microsoft calls it "the new next generation OpenAI Large Language Model" which is "faster, more accurate, and more powerful" than ChatGPT and GPT-3.5.

Of course, considering that Microsoft has invested heavily in OpenAI, it's not surprising that it has access to technology beyond the purview of ordinary developers, and it also uses its own technology in Bing.

Of course, OpenAI is also constantly improving the public ChatGPT model, and developers also have the flexibility to choose specific versions, such as gpt-3.5-turbo-0301 released today. It will be supported until at least June 1st, with gpt-3.5-turbo updated to a stable version in April.

According to reports, OpenAI will provide 1,000 tokens at a price of $0.002, “which is 90% cheaper than our existing GPT-3.5 model. ”, in part due to “a series of system-wide optimizations.”

##▲Request

##▲Response

▲ Binding Python

Although 1000 tokens may sound like a lot, it may take several tokens to normally send a text fragment. (The "Token" mentioned here refers to the text chunks that the system breaks down sentences and words into in order to predict what text should be output next.)

According to the official OpenAI documentation, the group "ChatGPT is great!" A word requires six tokens - its API breaks it down into "Chat", "G", "PT", "is", "great" and "!".

Of course, the company also provides developers with a tool to check how many tokens are needed to interpret a string of text. As a general rule of thumb, they say, "a token usually corresponds to about 4 characters" in English.

Developers can also get a dedicated ChatGPT instance if they run large amounts of data through the API, the company said. Doing so supposedly gives you greater control over parameters like how long it takes to respond to a request and how long a conversation with the bot can last.

The above is the detailed content of OpenAI announces open API so developers can integrate ChatGPT into their products. For more information, please follow other related articles on the PHP Chinese website!