OpenAI has announced the release of its new ChatGPT and Whisper APIs, allowing developers to build apps and products with the power of artificial intelligence. The ChatGPT API leverages the gpt-3.5-turbo model, also used in OpenAI’s ChatGPT product, and is suitable for a range of non-chat use cases. The company claims that early testers of the Chat API have migrated from text-davinci-003 to gpt-3.5-turbo with only a slight adjustment needed to their prompts. The Chat API is priced at $0.002 per 1K tokens, which is 10 times cheaper than existing GPT-3.5 models.
OpenAI also announced that it is constantly improving its Chat models and will make these improvements available to developers as well. Developers who use the gpt-3.5-turbo model will always receive the recommended stable model while still having the flexibility to opt for a specific model version. For example, OpenAI is releasing “gpt-3.5-turbo-0301” today, which will be supported through at least June 1st, and it will update gpt-3.5-turbo to a new stable release in April.
The Whisper API, OpenAI’s speech-to-text model, is also available through the API as whisper-1 starting today. OpenAI claims that it is the fastest, cheapest, and most convenient way to use the model. The API is priced at $0.006 per minute, rounded up to the nearest second.
OpenAI has also made some policy changes in response to feedback from API customers. The company will no longer use data submitted through the API for model training or other service improvements unless users were explicitly opt-in. API users’ default data retention policy is now 30 days, with options for shorter retention windows depending on user needs. The company has also simplified its Terms of Service and Usage Policies, including terms around data ownership where users own the input and output of the models.
Snap, Shopify, and Instacart are among OpenAI’s early partners who have already built next-generation apps powered by ChatGPT and Whisper APIs. Developers can read more about the APIs and early partners’ use cases in OpenAI’s documentation.