OpenAI has just released announced that developers can now integrate the ChatGPT and Whisper models into their apps and products via the API. This means you now have access to advanced language and speech-to-text capabilities powered by these models.
ChatGPT API for cutting-edge language processing
ChatGPT models offer not only chat capabilities, but also advanced language processing. Through a series of system optimizations, they have achieved 90% cost savings since December, and these savings are now being passed on to API users.
Developers can now use the open-source Whisper large-v2 model with much faster and cost-effective results. Users of the ChatGPT API can expect continuous model improvements and choose the option for dedicated capability for deeper control over the models.
Examples of ChatGPT and Whisper
In OpenAI’s news release, they mention several apps already using the new ChatGPT and Whisper APIs. For example, Snapchat has introduced My AI, an experimental feature running on the ChatGPT API that offers users a friendly, customizable chatbot with recommendations and can even write a haiku for friends in seconds.
Quizlet is also introducing Q-Chat, a fully adaptive AI tutor that engages students in adaptive questions based on relevant study materials delivered through a fun chatting experience.
Instacart allows customers to ask about food and get inspiring, for-sale answers through Ask Instacart. This uses the ChatGPT API combined with Instacart’s own AI and product data from their 75,000+ retail partners.
Shopify’s Shop app uses the ChatGPT API to power the new shopping assistant, which makes personalized recommendations based on buyer requests.
And the AI-powered language learning app Speak is already using the Whisper API to develop the latest AI speaking companion, aimed at building the best path to spoken fluency for language learners of any level.
ChatGPT API pricing and upgrades
The price for the ChatGPT model, gpt-3.5-turbo, is the same model used in the ChatGPT product. It costs $0.002 per 1k tokens, which is 10x cheaper than existing GPT-3.5 models. Developers using the gpt-3.5 turbo always get the recommended stable model, while still having the flexibility to choose a specific model version.
Dedicated instances and Whisper API
In addition to the ChatGPT API, OpenAI now offers dedicated instances for users who want more control over the specific model version and system performance. By default, requests are executed on infrastructure shared with other users, who pay per request. With dedicated instances, developers pay for a period of time for an allocation of compute infrastructure reserved to serve their requests.
Developers have full control over instance load (higher load improves throughput but makes each request slower), the ability to enable features such as longer context limits and the ability to pin the model snapshot. Dedicated instances can make economic sense for developers running more than ~450 million tokens per day. Moreover, directly optimizing a developer’s workload against hardware performance makes costs significantly lower compared to shared infrastructure. For questions about dedicated instances, please contact OpenAI.
The Whisper API uses the speech-to-text model Whisper, which opened in September 2022. The large-v2 version of this model is now available through the API and provides convenient on-demand access at a cost of $0.006 / minute. In addition, the highly optimized server stack provides faster performance compared to other services.
Whisper API is available through OpenAI’s transcriptions (transcribes in the source language) or translations (transcribes to English) endpoints and accepts various formats (m4a, mp3, mp4, mpeg, mpga, wav, webm).
Conclusion
With the ChatGPT and Whisper APIs, OpenAI offers developers access to cutting-edge language and speech-to-text capabilities. The APIs have already powered several applications, including Snapchat’s My AI, Quizlet’s Q-Chat and Instacart’s Ask Instacart. Dedicated instances offer developers greater control over specific model version and system performance, while the Whisper API provides convenient access to speech-to-text capabilities at an affordable price.
