OpenAI has launched a way for phone users to call ChatGPT for free for up to 15 minutes by dialing 1-800-CHATGPT. Kevin Weil, chief product officer of OpenAI, said during the live broadcast that this feature was a project that was launched just a few weeks ago. Users can now call ChatGPT in the United States and send messages via WhatsApp globally at 1-800-242-8478. Each phone number has a monthly calling allowance of 15 minutes, and users can even set up several Google Voice numbers to get as much calling time as possible.

The phone number was created using OpenAI’s real-time API, while the WhatsApp functionality is provided by GPT-4omini through integration with the WhatsApp API.


OpenAI sees this feature as an important stepping stone for those new to artificial intelligence because the service is a simplified version of ChatGPT compared to the web-based ChatGPT and provides a "low-cost way to try it through familiar channels." The company noted that existing users looking for more comprehensive features, higher usage limits and personalization options should continue to use their regular ChatGPT accounts through traditional channels.

Interestingly, Google launched a similar tool called GOOG-411 in 2007 to provide free voice directory assistance. But some have speculated that the service was shut down because the company had achieved its basic goal: amassing a database of enough voice samples to advance its speech recognition technology.

At the time, Google Vice President Marissa Mayer put it bluntly: "Our speech recognition experts said: If you want us to build a really powerful speech model, we need a lot of phonemes, which are syllables made by specific sounds with specific intonations. So we need a lot of people speaking words, say something so that we can ultimately train on that. ... So 1-800-GOOG-411 is about that: getting a lot of different speech samples so that when you make a call or we try to get speech from a video, we can do it with high accuracy."

OpenAI spokesperson Taya Christianson said the company will not use the calls to train large language models.