OpenAI made its Decisions API model widely available to users for the first time, jumping straight into the fast-growing market for quick and cheap intelligence tools. This space recently turned TypeSafe into a giant valued at $7.6 billion, following the success of its Jev model.
Key takeaways
- OpenAI rolled out its Decisions API to all customers after initially announcing it on September 29.
- The system speeds up inference times significantly by reconfiguring ChatGPT Luna instead of building a new model from scratch.
- Early testers like Box, Datadog, Ramp, and Speak used the software to replace older technology with higher throughput.
- Competitors like TypeSafe and Liquid continue releasing new versions and benchmarks for quick classification tools.
What is the OpenAI Decisions API?
The Decisions API acts as a super classifier that can quickly pick from a set of options while providing a confidence score. Nikunj Handa, the product lead for OpenAI APIs, told TechCrunch that his team saw similar offerings and rushed to build their own version by tweaking ChatGPT Luna to cut costs and speed up tasks.
Initial trials let software developers swap out standard OpenAI models to speed up workflows. According to TechCrunch, Speak CTO Andrew Hsu noted that his team cut down action times from 1.4 seconds to .13 seconds using the new system. OpenAI also used the product to boost the responsiveness of its Dots voice assistant.
How do different Decision model choices compare?
Public rankings show mixed results across various platforms. Benchmarks like the AIM System One Index and the Artificial Analysis Decision Index place Liquid's D1 Ultra and Base Lab's Elephant-1 near the top. Meanwhile, OpenAI's entry sits further down those lists, though TypeSafe's offerings remain the most queried on OpenRouter.
Developers also test these tools using fun challenges. Opper AI created a test called JevMan where models play PacMan to test rapid choices, and the newest Jev model leads that specific ranking. Handa told TechCrunch that OpenAI's tool holds an advantage in handling images, while customers want even more flexibility for visual outputs.
What happens next for fast AI tools?
As these tools become faster and cheaper, developers expect to see a wave of new workflows emerge. Handa pointed out that pushing speeds down to 60 milliseconds could open the door to analyzing every single frame in a video stream. Companies are targeting better regional latency and adding voice and visual modalities to support these expanding tasks.
Frequently asked questions
It is a model that acts as a super classifier, letting users pick from options quickly with a confidence rating while operating faster and cheaper than standard intelligence tools.
OpenAI announced the product at its developer day on September 29 and made it available to all customers on Thursday.
Rather than building a specifically calibrated system from scratch, OpenAI took ChatGPT Luna and reconfigured it to lower costs and speed up inference times.
Early trial customers included Box, Datadog, Ramp, and Speak, who used the API to boost throughput and cut down response times.
JevMan is a benchmark created by Opper AI where decision models play PacMan to test their ability to make quick, high-stakes choices within hundreds of milliseconds.
Based on reporting by TechCrunch. Read the original article
