OpenAI has launched its Decisions API, making the model available to all customers following its initial announcement on September 29. The tool is designed to provide rapid decision-making capabilities, with inference speeds reduced to between 100 and 180 milliseconds, compared to the previous one to three seconds.
The API is a reconfigured version of the ChatGPT Luna model, optimized for lower costs and higher throughput. OpenAI has integrated new voice and visual modalities into the public release and improved latency by deploying inference resources across various geographic regions.
Early adopters, including Box, Datadog, Ramp, and Speak, have utilized the API to replace standard large language models. Speak reported that the integration reduced action response times from 1.4 seconds to 0.13 seconds. OpenAI also implemented the technology to enhance the responsiveness of its Dots voice assistant.
The decision model market currently features competitors such as TypeSafe, which is valued at $7.6 billion, and Liquid, which recently released its D1 Ultra model. While OpenAI’s model currently ranks eighth and fourteenth on the AIM System One and Artificial Analysis indices, the company is focusing on multi-modal flexibility and potential future speed improvements.
