Key Takeaways
- OpenAI has released the Decisions API, making decisions up to 10 times faster.
- The API supports probability estimates, choices between options, and scores based on predefined criteria.
- Early customers report significant improvements in response times.
OpenAI has launched the Decisions API, a new tool designed to make decisions up to 10 times faster than its traditional Responses API. This development is aimed at applications requiring near-instant responses, such as financial trading or real-time customer service.
According to OpenAI, the Decisions API focuses on selecting answers from predefined options, making it suitable for tasks where quick decisions are crucial. The API supports three types of responses: probability estimates, choices between options, and scores based on predefined criteria.
Nikunj Handa, OpenAI’s API product lead, told TechCrunch that the company has reduced decision-making time from one to three seconds to approximately 100–180 milliseconds. This significant improvement is achieved by modifying the existing language model to reduce processing time and costs, rather than training a separate model.
Performance and pricing details are outlined in the following table. Traditional AI response time is 1–3 seconds, while the Decisions API response time is 100–180 milliseconds. Input pricing is $0.10 per million tokens, with cached input pricing at $0.01 per million tokens. Output token charges are none, and the maximum number of questions per request is 400. Supported inputs include text, images, and structured data.
Early adopters of the Decisions API include Box, Datadog, Ramp, and Speak. Andrew Hsu, CTO of Speak, reported that the language-learning platform reduced the time between AI actions from 1.4 seconds to 0.13 seconds, representing an improvement of more than 90%.
With the release of the Decisions API, OpenAI enters competition with TypeSafe, the startup behind the Jev decision-making model, which recently reached a valuation of approximately $7.6 billion. Other competitors include Liquid and Base Lab, which have introduced competing models. The following table compares their positions on two public benchmarks, based on the reported rankings.
AI Model | AIM System One Index | Artificial Analysis Decision Index
TypeSafe Jev | 7th | 3rd
OpenAI Decision Model | 8th | 14th
Liquid D1 Ultra | Strong performer | Strong performer
Base Lab Elephant-1 | Strong performer | Strong performer
Handa suggested that further improvements could reduce response times to approximately 60 milliseconds, potentially allowing developers to make AI decisions for individual frames in videos.
The growing competition highlights increasing demand for AI systems that can handle repetitive decisions without the higher costs and processing times of traditional language models.
The company reduced decision-making time from one to three seconds to approximately 100–180 milliseconds.
Nikunj Handa, OpenAI’s API product lead
