OpenAI Launches ChatGPT API, Economic Version Of AI Chatbot For App Developers

OpenAI launched the Application Programming Interface (API) for ChatGPT, (photo: Jonathan Kemper / Unsplash)

JAKARTA - OpenAI launched the Application Programming Interface (API) for ChatGPT, making it easier for third parties or any business to build ChatGPT technology into their applications, websites, products, and services.

The chatbot API is powered by GPT-3.5 turbo, the same model used for ChatGPT web products. Developers can easily interact with it through simple endpoints and use it for any tasks ChatGPT can perform, from within their websites or apps.

This model runs on Microsoft Azure computing infrastructure that connects to the user's endpoint. This can make it run independently of server loads on the ChatGPT website, offering a dedicated lane business for AI processing.

However, the most important thing about the ChatGPT API is the cost. AI-powered chatbots usually require data centers to run, forcing providers to spend a lot of money on hardware and electricity to serve millions of users.

But through software optimization, OpenAI has managed to reduce costs to run ChatGPT by up to 90 percent.

In particular, OpenAI plans to collect US$0.002 API users (Rp30 thousand) for each 1,000 tokens, or 750 words, generated by ChatGPT.

The fee is 10 times cheaper than the company's API access to the previous large language model Daviinci GPT-3.5, which can also support chatbot services.

OpenAI said this version of the API may be the most economical for developers who expect more than 450 million tokens per day in demand, and can be approved through direct contact with the company.

Furthermore, data processed through API is not used for model training or other improvements to its services unless the organization chooses to participate.

OpenAI has also suspended its pre-launch review policy requiring developers to tag what they used on the model before it could be integrated into their app.

"Data sent to the OpenAI API are not used for training, and we have a new 30-day retention policy open to fewer cases per case," OpenAI CEO Sam Altman tweeted, as quoted by ITPro, Friday, March 3.

"We have also removed our pre-launch reviews and made the terms of our service and use policies more friendly to developers," he added.

The latest model, gpt-3.5-turbo-0301, will be supported until June and a new stable release of gpt-3.5-turbo is expected in April. Developers will be given the option to adopt a stable model or special model according to their needs.