OpenAI has launched its Decisions API model for broad public use, challenging the market dominance of TypeSafe’s Jev, a $7.6 billion startup that pioneered fast and cost-effective decision-making AI models. Unlike Jev, which was designed specifically for decision tasks, OpenAI adapted its existing ChatGPT Luna model to deliver rapid responses with significantly reduced latency, achieving inference speeds between 100 to 180 milliseconds. OpenAI revealed this capability during its developer day on September 29, and has since rolled out the API to customers worldwide, incorporating voice and visual abilities and optimizing server locations to reduce delays.

Early adopters of OpenAI’s Decisions API, including companies like Box, Datadog, Ramp, and Speak, report notable improvements in task throughput and responsiveness, with Speak’s CTO highlighting a drastic reduction in response time from 1.4 seconds to just 0.13 seconds. OpenAI’s decision model also powers enhancements in its own Dots voice assistant. Despite these promising performance gains, the ideal decision model for specific applications remains uncertain, with variability in accuracy, cost, and speed demands influencing user preferences. TypeSafe, meanwhile, continues refining Jev with new capabilities such as image handling and conditional logic, maintaining strong popularity in the AI ecosystem.

Competition intensifies as industry benchmarks yield varied results: TypeSafe’s Jev often ranks high in decision model indices, while OpenAI’s model scores moderately but offers advantages in multimodal processing, especially image classification through Luna. Other contenders, such as Liquid’s D1 Ultra and Base Lab’s Elephant-1, also gain recognition, highlighting a diverse and rapidly evolving decision AI landscape. Experts suggest ongoing advancements could push decision models toward ultra-low latency, enabling real-time applications like per-frame video analysis.

OpenAI’s decision model expansion reflects a broader trend of AI systems becoming more specialized and integrated into practical workflows, unlocking "hidden" use cases by enabling faster, cheaper, and more reliable decision-making. Nikunj Handa, product lead for OpenAI’s APIs, envisions continued speed improvements and tailored training to enhance decision performance further. As decision AI technology matures, it may follow the Jevons paradox, increasing efficiency and proliferation due to decreased resource costs, ultimately transforming numerous industries reliant on swift intelligent automation.