OpenAI made its Decisions API available to all customers on Thursday, October 22, providing a cheaper and faster method for developers to use artificial intelligence on tasks requiring quick choices. The firm initially introduced this technology at a developer event on September 29. According to OpenAI, the Decisions API runs up to 10 times quicker than the standard Responses API, making it ideal for software needing near-instant feedback.
How the Decisions API Works
Instead of generating detailed answers like standard large language models, the Decisions API picks solutions from predefined lists. Developers supply questions alongside possible answers so the model can classify data, pick actions, or compute confidence metrics.
The system handles three response formats, which include probability estimates, choices between options, and scores based on set criteria. Rather than training a completely separate model, OpenAI modified an existing language model to cut down expenses and processing time.
Nikunj Handa, who serves as OpenAI's API product lead, told TechCrunch that the organization brought response times down from one to three seconds to roughly 100-180 milliseconds.
Performance and Pricing
OpenAI launched the Decisions API to lower processing costs and improve response speeds. Nikunj Handa noted that decision-making time dropped from one to three seconds to about 100-180 milliseconds.
Input pricing is set at $0.10 per million tokens, while cached input costs $0.01 per million tokens. Output token charges do not apply, and requests can include up to 400 questions using text, images, and structured data as supported inputs.
Early users of the technology include Box, Datadog, Ramp, and Speak. Speak CTO Andrew Hsu reported that the language-learning platform cut the time between artificial intelligence actions from 1.4 seconds to 0.13 seconds, which is an improvement exceeding 90%.
Competition With TypeSafe
This rollout pits OpenAI against TypeSafe, which is the startup behind the Jev decision-making model. TypeSafe holds a valuation of roughly $7.6 billion, and rivals like Liquid and Base Lab also offer competing models.
Liquid D1 Ultra and Base Lab Elephant-1 did not receive exact numerical rankings because precise positions were not provided, and rankings can shift when benchmarks receive updates.
AI Models Could Become Even Faster
OpenAI anticipates that future enhancements might lower response times to about 60 milliseconds. Handa indicated that such speeds could eventually let developers run artificial intelligence decisions for individual video frames.
At the same time, TypeSafe added conditional decision-making capabilities and image support to Jev. This rising competition shows a growing need for artificial intelligence tools capable of managing repetitive decisions without the elevated costs and delays of standard language models.
OpenAI expects these affordable, quick decision models to power additional applications, such as automated workflows, interactive software, real-time voice assistants, and image classification.
