OpenAI Releases Decisions API to All Customers, Targeting Fast, Cheap 'Decision Model' Market
The Facts
- OpenAI made its Decisions API generally available to all customers on Thursday, after announcing it at its developer day on September 29. TechCrunch only
- The Decisions API is built on ChatGPT Luna, reconfigured to speed up inference and lower cost rather than a specifically trained decision model, according to OpenAI product lead Nikunj Handa. TechCrunch only
- Handa said the API cuts response times from one to three seconds to roughly 100-180 milliseconds. TechCrunch only
- The public version adds voice and visual modalities, and OpenAI targeted inference resources geographically to lower latency in regions worldwide. TechCrunch only
- Early customers included Box, Datadog, Ramp and Speak. Speak CTO Andrew Hsu said the API cut the company's time between actions from 1.4 seconds to 0.13 seconds. TechCrunch only
Dive deeper · 4 more factsShow less
- OpenAI also used the product to improve the responsiveness of its Dots voice assistant. TechCrunch only
- The product competes with TypeSafe's Jev model. TypeSafe, founded by a former OpenAI researcher, was recently valued at $7.6 billion. TechCrunch only
- On the AIM System One Index and the Artificial Analysis Decision Index, the latest Jev model ranks seventh and third, while OpenAI's decision model ranks eighth and fourteenth. Liquid's D1 Ultra and Base Lab's Elephant-1 tend to be favored on these indices. TechCrunch only
- TypeSafe's models are currently the most widely queried decision models on OpenRouter, and the latest Jev model tops the JevMan PacMan benchmark built by Opper AI. TechCrunch only
Context
OpenAI has opened its Decisions API to all customers, entering a category of fast, low-cost classifier-style models that TypeSafe's Jev established. These models choose among options and return a confidence rating, and developers have used them in place of larger language models in some workflows.
OpenAI's version repurposes its existing ChatGPT Luna model instead of using purpose-built training, though Handa said specific training may follow. Benchmark results vary by index and workload, and TypeSafe's CEO has urged users to test models on their own tasks.
How the Agent Panel Scored This
Independent AI personas each scored importance 1–10; the story's score is a trimmed mean (highest and lowest votes dropped). Vote spread: 1.
| 4 | Policy Expert | OpenAI's general availability of a commercial API is a concrete product launch with some industry impact, but it is a single-company product release from a single source with limited broad societal effect, so it falls in the minor-to-moderate range. |
| 4 | Business Analyst | OpenAI's general availability of a new API product is a concrete product launch with some competitive implications in a niche AI-model category, but it has limited broad economic impact and the story is a single-source product announcement rather than market-moving data or a major corporate or regulatory action. |
| 4 | News Editor | OpenAI's general availability of a niche developer API for a specialized model category is a concrete product launch of moderate-to-limited impact, mostly relevant to developers and the AI tooling market rather than the broad public. |
| 4 | Global Affairs Analyst | OpenAI's general availability of a commercial API is a concrete product launch with moderate competitive significance in the AI market, but it has limited geopolitical or broad societal impact. |
| 4 | Critical Reader | OpenAI's general availability of a niche API product is a concrete launch with measurable latency gains, but it is a commercial product release in a specialized market with limited immediate impact on most people and benchmark results are mixed. |
| 3 | Science & Tech Analyst | A commercial product general availability from OpenAI in a niche model category, with speed gains and competitive benchmark positioning but no peer-reviewed breakthrough or broad public impact. |
| 3 | Informed Citizen | A product general-availability release from one company in a niche developer API category, with limited immediate impact on ordinary people and no broad consumer effect expected within a week. |