After a limited test period, Openai provided the developers with its CHATGPT and Whisper models, which can now integrate interaction with the chat boot and convert speech into the text into their own applications using the API.
The new family of ChatGPT models is called “GPT-3.5-Turbo” and is offered at a price of $ 0.002 for 1000 tokens (text blocks that the system divides sentences and words. In this case, these are approximately 750 words), which, according to Openai , 10 times cheaper than previous GPT-3.5 models.
The cost of the Whisper “Large-V2” is $ 0.006 per minute. There is also version Whisper with an open source, but Openai admits that it may be difficult to run.
The data processing specialist Max Wolf noted that the prices for the API Openai are extremely low. Probably, the company will suffer record losses to cut off competitors and then overtake them. However, there are no guarantees that these prices will not rise later.
Access to ChatGPT using a Curl request looks like this:
Openai promised to make changes based on the reviews of the first users. Changes include:
- termination of the use of data sent through the API;
- Add to the conditions of service and the policy of using the point that “users own all the introductory and output of these models”;
- and others adjustments.
Openai provides the first customers with access to these two models through the API. As a result, many applications based on Openai neural networks appeared, including:
- my ai, a virtual assistant for subscribers Snapchat+;
- shop, Shopify platform assistant based on Chatgpt;
- Speak, Language Learning Application, based on Whisper.
Whisper is a speech recognition system issued in September 2022. A neural network can convert speech into a text, which can then be entered in ChatGPT or use in any other application. Whisper, for example, can be used in decoding interviews.