OpenAI announces Decisions API, which returns predefined answers with confidence scores, similar to TypeSafe's Jev, in limited preview with no pricing details
OpenAI had already broadened its Responses API for agent-building and added remote MCP, image-generation and Code Interpreter support in 2025. The Decisions API separates a narrower task—returning from predefined answers with confidence scores—into its own limited-preview endpoint, extending OpenAI's push from general models toward application-facing building blocks.
The move also follows OpenAI's 2024 expansion of enterprise API controls and Assistants capabilities, making confidence-bearing outputs more relevant to teams that need an explicit signal alongside a model-selected result.
First-order effects
OpenAI API customers in the limited preview can build flows around predefined decisions and their confidence scores rather than deriving that signal themselves.
TypeSafe's Jev gains a much larger platform rival for decision-model workloads, while OpenAI has not disclosed pricing for a direct cost comparison.
Second-order effects
Developers evaluating Jev and OpenAI will need to benchmark answer quality, confidence calibration and image handling, an approach reflected in discussion around JevBench.
Procurement teams cannot yet calculate the cost per decision, leaving performance and integration fit—not price—as the immediate basis for comparison.
Third-order effects
If developers adopt confidence-scored decision endpoints, model APIs may compete less on open-ended generation alone and more on task-specific interfaces with measurable outputs.
The pattern would favor API platforms that package models with workflow tooling, raising the pressure on specialist decision-model providers to differentiate on benchmarks, cost or deployment fit.
The trend: AI application platforms are moving from general-purpose model access toward workflow-native endpoints that expose structured, evaluable outputs for specific tasks.
It was obvious, but f*ck that happened fast. Ok, so now I'll have to benchmark it, and I'll do so pretty fast. If anyone at @OpenAI can give me early/free access, to put it on JevBench, it would be appreciated. /cc @sama
OpenAI (and probably Anthropic and DeepMind) coming after Jev with their own decision model (aka System One Model) was an obvious and predictable outcome. However, this doesn't necessarily mean that Jev will be eliminated, because: 1) It was created by one of the former crack eng…
curious if this is just an API wrapped on Luna? Luna is VERY cheap, but not Jev cheap Image inputs is great though, I've had it process images for $0.001
Decisions API, for lightning fast constrained decision making powered by Luna. Supports visual inputs, and tuned to be able to make decisions in less than a few hundreds of milliseconds end to end.
New decisions API from OpenAI powered by giving Luna model a predefined set of options to choose from! Looks really good. Congrats to the Jev team for creating a new category too😉