OpenAI Launches Decisions API Globally at 100ms to Rival TypeSafe Jev

Executive Overview

OpenAI has officially made its Decisions API available to all customers globally as of October 22, 2026, following its initial reveal at the developer day on September 29. The model, built on a reconfigured version of ChatGPT Luna, dramatically reduces response latency from 1-3 seconds to 100-180 milliseconds and introduces new voice and image support. This launch positions OpenAI in direct competition with TypeSafe's Jev model, which is valued at $7.6 billion, signaling the emergence of a market for fast, low-cost classification models.

Deep-Dive Features & Architecture

Nikonj Handa, product manager for OpenAI interfaces, revealed that the team moved quickly after seeing TypeSafe's Jev model, opting to reconfigure ChatGPT Luna rather than build a calibrated decision model from scratch. "The big unlock for Decisions API was speed, cutting things from one to three seconds down to 100, 120, and 180 milliseconds," he said. He added that the company may train custom models in the future, but "for now, the old regular Luna is what gets the job done."

In the general release, OpenAI added new voice and visual modalities and worked on distributing inference resources geographically to ensure low latency across different world regions. Companies like Box, Datadog, Ramp, and Speak tested the API to replace LLMs with higher productivity; Andrew Hsu, CTO of Speak, said: "Decisions API allowed us to cut the time between actions from 1.4 seconds to 0.13 seconds." OpenAI also used the product to improve the response of its voice assistant Dots.

OpenAI's most prominent competitive advantage is multimodality; Handa confirmed that Luna is "very strong at image classification and shows better calibration," and that customers are requesting greater flexibility in using images as outputs. The team is aiming for speeds as low as 60 milliseconds to enable use cases like calling an API per video frame.

Benchmark & Competitive Performance

On the AIM System One Index, the OpenAI Decisions Model ranked 8th, and on the Artificial Analysis Decision Index it ranked 14th. Meanwhile, TypeSafe's latest Jev model ranked 7th and 3rd respectively, and Liquid's D1 Ultra and Base Lab's Elephant-1 topped most general benchmarks. Nevertheless, TypeSafe's offerings remain the most queried on OpenRouter. In the JevMan benchmark, where decision models play PacMan, the latest Jev model tops the list, while OpenAI's ranking was not mentioned.

Industry Impact & Enterprise Adoption

The Decisions API represents a significant opportunity for developers to build ultra-fast interactive applications at a fraction of the cost of traditional LLMs, particularly in areas such as voice assistants, automated customer service, and content classification. Because the model runs on multilingual Luna, Arabic token efficiency depends on model calibration, and developers are advised to test actual costs on Arabic text before production. The new image and voice support further expands the possibilities for regional applications, enabling richer, multimodal user experiences.

Conclusion

OpenAI's global rollout of the Decisions API marks a strategic shift toward specialized, high-speed classification models, directly challenging TypeSafe's Jev. With latency reductions of up to 90% and multimodal capabilities, the API opens new doors for developers worldwide, including in the Arabic-speaking market. While benchmark rankings show room for improvement, the combination of speed, cost, and versatility positions OpenAI as a formidable player in the evolving decision model landscape.

Media Source: TechCrunch AI | البيان الرسمي للشركة: المصدر الأصلي | Fact Verification & Analysis: AI Tools Oasis