OpenAI makes its Jev competitor widely available as decision model competition heats up
TechCrunch Tim Fernholz ● Covered by 2 sources
OpenAI just opened up its fast Decisions API to everyone. It’s chasing the cheap, quick model market that TypeSafe turned into a $7.6 billion business.
Based on reporting by TechCrunch, Tim Fernholz — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
OpenAI has pushed its Decisions API model out to all customers, stepping straight into the scramble for fast, cheap AI that TypeSafe helped popularize with its Jev model. The pitch is simple: take a set of options, make a quick call, and return a confidence score. For developers, that can be enough to replace a larger LLM in some workflows, with a lot less cost and a lot less waiting.
Nikunj Handa, who leads product for OpenAI’s APIs, said the company saw Jev and immediately spotted “a million use cases.” OpenAI didn’t build a fresh decision model for this release. It took ChatGPT Luna and reworked it to cut inference time and lower cost. Handa said the big shift has been speed, moving from one to three seconds down to roughly 100 to 180 milliseconds. He also said OpenAI may eventually train models specifically for this job, but for now Luna is doing the work.
The company announced the feature at its developer day on September 29, then made it broadly available on Thursday. The public version now includes new voice and visual modes, and OpenAI says it has spread inference resources across regions to reduce latency. Early tests with Box, Datadog, Ramp, and Speak showed engineers swapping in the Decisions API for OpenAI’s larger models when they wanted higher throughput.
Speak’s CTO, Andrew Hsu, said the time between actions dropped from 1.4 seconds to 0.13 seconds. OpenAI has also used the product to make its Dots voice assistant respond faster. But the field is still messy. Benchmarks disagree on what “good” even means, and TypeSafe’s founder argues that real-world testing matters more than charts. Even so, the benchmark race is now part of the product story, with OpenAI’s model placing eighth and fourteenth on two public indices while TypeSafe’s latest Jev remains more widely queried on OpenRouter.
The one place OpenAI seems especially confident is multimodality. Handa said Luna is strong at image classification and that customers want the Decisions API to handle images more flexibly, including as outputs. He also floated an even faster future, with calls around 60 milliseconds and maybe one per video frame. That is the real signal here: the market is moving from clever demo to infrastructure, and everybody wants their model to be the one doing the split-second deciding.
My take — AI-written commentary, not fact-checked reporting
This is the usual AI market move: one company proves a niche is real, and five others rush in to claim they were obviously thinking about it all along. The interesting part isn’t the benchmarks; it’s how quickly “fast enough and cheap enough” becomes a product category. That’s where the money is, and where the hype starts pretending to be engineering.
Read more about this at: TechCrunch
