OpenAI Opens ChatGPT and Whisper APIs, Ushering in a Mass AIGC Era

On March 2, OpenAI announced via its official blog that it is opening the ChatGPT and Whisper model APIs, allowing developers to integrate them into applications and products.

The ChatGPT API provides access to the same gpt-3.5-turbo model used by ChatGPT, priced at $0.002 per 1,000 tokens, roughly ten times cheaper than the older GPT‑3.5 model.

OpenAI cites two reasons for the release: first, continuous optimization and innovation —by enabling more enterprises and developers to use gpt-3.5-turbo, the company hopes to discover issues, improve the model, and drive new feature development; second, benefiting humanity —the cost of ChatGPT has been reduced by 90% since December 2022, and the saved funds can be passed to developers, while the model is positioned as a shared human resource.

This launch dramatically lowers the technical and financial barriers for enterprises, AIGC startups, and individual developers, enabling use cases such as automatic email generation, Python code writing, and intelligent customer‑service bots without years of R&D or billions of dollars of investment.

For large enterprises that process more than 450 M tokens per day, OpenAI offers customized services to meet higher efficiency, load, and throughput requirements.

The Whisper API now offers the large‑v2 model, billed at $0.006 per minute, supporting multiple audio formats (m4a, mp3, mp4, mpeg, mpga, wav, webm) and 98 languages, and can be accessed via the transcriptions (source‑language transcription) or translations (English translation) endpoints.

Based on six months of feedback, OpenAI has made specific policy changes: data submitted through the API is no longer used for model training unless the organization opts in; a default 30‑day data retention period is applied with optional longer retention; pre‑release review has been removed; developer documentation has been optimized; and the service terms now state that users own both the inputs and outputs of the model.

Early adopters illustrate practical applications: Quizlet, a global learning platform with over 60 million users, built “Q‑Chat,” an adaptive AI tutor using the ChatGPT API; Speak, a fast‑growing Korean English‑learning app, leverages Whisper for speech‑to‑text and translation, providing human‑like feedback to learners; and Shopify’s Shop app introduced a “smart guide” that uses the ChatGPT API to generate personalized product recommendations during shopping.

API‑submitted data is not used for model training unless the organization opts in.

Default 30‑day data retention with optional longer periods.

Removal of pre‑release review (now automated monitoring).

Improved developer documentation.

Simplified terms granting users ownership of model inputs and outputs.