AI
OpenAI broadly releases a low-cost Decisions API for fast classification and multimodal tasks
OpenAI has broadly released its Decisions API, a lower-latency model designed to choose among options with a confidence rating. The company says it reduced response times from roughly one to three seconds to as little as 100–180 milliseconds, while early customers reported faster application interactions. The API adds voice and visual inputs. Independent rankings place it behind several competitors, underscoring that performance depends heavily on the workload.
Editorial summary · Publisher article
What matters
- OpenAI reconfigured its Luna model for faster, cheaper inference rather than training a separate decision model.
- Customers including Speak, Box, Datadog and Ramp used it to increase throughput or reduce action delays.
- The public release includes voice and visual modalities and geographically targeted inference resources.
- Benchmarks rank competing models ahead of OpenAI’s offering on some decision tasks.
Context & caveats
- Latency and customer improvements are reported claims; model quality and calibration vary by application.
Why this made the edition
The report explains a significant AI infrastructure development with pricing, latency measurements, customer examples, competing benchmarks and limitations. Company claims are attributed, and the article acknowledges uncertainty about how decision models should be evaluated.
Source material