Sam Altman announced OpenAI’s new “Decisions API” during the company’s Dev Day event on Tuesday.
The tool mirrors Jev, a model from TypeSafe AI released earlier this month for software automation. Jev acts as a classifier built on a large language model. Developers provide it with a set of choices, and it outputs probabilities for each option quickly and cheaply.
OpenAI’s offering appears to be the same product. At the event, Altman described the API as a method to give the lab’s Luna model a predefined list of options. These could include categories for image classification or different agent behaviours.
“By focusing the model on that choice, we can make it extremely fast while keeping capabilities like image understanding, broad language support, and safety protections,” Altman said.
TypeSafe did not answer questions from TechCrunch regarding the new product. However, CEO Diogo Almeida, a former OpenAI engineer who co-invented reinforcement learning, joked on X about the start of clone wars.
He added that OpenAI’s interest could be “a sign…that building in a System One compatible way is the future.” System One is TypeSafe’s term for fast, intuitive thinking, while System Two covers deliberate reasoning.
The subtext suggests large language models are not the right solution for much software because they are comparatively slow and expensive. Developers have used Jev to augment LLMs and found the process faster and cheaper.
It is unclear how similar Decisions API will be to Jev. OpenAI released it as a limited preview, and TechCrunch has not yet seen developers run it through its paces. Interest is evident from conversations on X.
Decisions API is not the only Jev-like API on the internet. Other startups are rolling out similar models, and OpenAI will not be the last tech giant to produce one. A key question is how well calibrated each of these decision models’ outputs will be to real life.
Almeida says his company’s moat is the synthetic data it creates to generate statistically useful outputs.
“Fast and cheap is very easy, you know,” Almeida told TechCrunch last week. “If you want it really fast and cheap, use dice, right? Intelligence is the hard part, and my North Star is always pushing the intelligence-per-dollar Pareto curve.”
After just weeks, it seems clear that these models have a future ahead of them. One likely application is monitoring and securing AI agents.
One of OpenAI’s new security measures following a series of incidents where its agents misbehaved on the open internet is using a separate model to watch for bad actions at significant compute cost.
Shapor Naghibzadeh, a long-time cybersecurity professional who leads the start-up QueryStory, thinks that a model like Jev could make that possible far more cheaply.
He built a demo for a hackathon held last weekend that uses Jev to check each agentic action against the task it was given. The system blocks actions it had high confidence were bad, flags others for review, and permits the rest.
In theory, such monitoring could have stopped the Hugging Face incident. Monitoring of that kind costs $2.94 with Jev, versus $372 with a frontier LLM.
A key observation is that Jev is arguably cheap enough to run on every agentic action. This offers a layer of review that could improve the reliability of agents writ large. It is the kind of thing TypeSafe was hoping to achieve, and now OpenAI has seen the value as well.
What it means
For builders of automated systems, the shift is practical. Monitoring agents for errors no longer requires expensive, slow models. Teams can now check every action against a predefined rule set at a fraction of the cost, making safety checks viable for high-volume workflows.
