OpenAI has opened its Decisions API to every customer, giving businesses a fast, cheap way to replace full language-model calls with a classifier that answers in 100 to 180 milliseconds. The general-availability launch happened Thursday, October 22, 2026, following the product’s first announcement at OpenAI’s developer day on September 29.
What happened
The Decisions API is OpenAI’s answer to TypeSafe’s Jev, the “decision model” that turned TypeSafe — a startup founded by a former OpenAI researcher — into a $7.6 billion company, TechCrunch reported. Decision models are AI’s “System One”: given a set of options, they quickly pick one and attach a confidence score, acting as a super-classifier rather than a full chatbot.
Rather than training a purpose-built model, OpenAI reconfigured its GPT-6 Luna model to speed up inference and lower cost. “The big unlock with Decisions API has been the speed, shrinking things from one to three seconds to 100, 120, 180 milliseconds,” Nikunj Handa, OpenAI’s product lead for APIs, told TechCrunch.
The public release adds voice and visual modalities and geographically targeted inference resources aimed at keeping latency low worldwide. In early trials, customers including Box, Datadog, Ramp, and Speak swapped the API in for OpenAI’s own LLMs. Speak, which builds AI language tutors, cut the time between actions from 1.4 seconds to 0.13 seconds, its CTO Andrew Hsu said. OpenAI also used the technology to sharpen the responsiveness of its own Dots voice assistant.
Competition is already fierce. Public benchmarks like the AIM System One Index and the Artificial Analysis Decision Index currently favor Liquid’s D1 Ultra and Base Lab’s Elephant-1, with the latest Jev model competitive and OpenAI’s Decision Model trailing in eighth and fourteenth place, respectively. TypeSafe remains the most-queried decision model on OpenRouter. Handa’s next target: 60-millisecond calls — “an API call for every single frame in a video.”
Why it matters for SMB
If your business runs a chatbot, support desk, or any automation that calls an AI model per interaction, you are likely paying full-model prices for simple questions: Is this spam? Which team should get this ticket? Which of these five replies fits best? A decision model does exactly that — and only that — far faster and at a dramatically lower cost, per TechCrunch’s reporting.
Latency also changes what feels possible. The 1.4-second-to-0.13-second jump Speak reported is the difference between a voice agent that feels sluggish and one that feels alive — a real-time experience small businesses could previously only dream of affording.
Keep one caveat in mind: decision-model quality varies by workload, and benchmarks currently put OpenAI’s model behind rivals like Jev. Test it on your own tasks before committing, as even TypeSafe’s CEO urges.
Pricing, availability, and what to do next
Availability. The Decisions API is generally available to all OpenAI API customers as of October 22, 2026, with text, voice, and visual inputs plus geo-targeted inference.
Pricing. OpenAI describes the cost as dramatically lower than full LLM calls for routing and selection work; price trackers, citing TechCrunch’s original reporting, put it at $0.10 per million input tokens. Check OpenAI’s pricing page for current rates before budgeting.
What to do next. Inventory the per-interaction AI calls in your product or workflows, pick one classification step — lead triage, ticket routing, intent detection — and run the Decisions API against your current model on a sample. Compare accuracy, speed, and cost before expanding.
For cost context, our AI tools pricing index tracks what popular business AI tools charge, and our n8n automation tutorial shows where a fast classifier could slot into a real small-business workflow.
Sources
- TechCrunch, “OpenAI makes its Jev competitor widely available as decision model competition heats up,” October 22, 2026: https://techcrunch.com/2026/10/22/openai-makes-its-jev-competitor-widely-available-as-decision-model-competition-heats-up/
- AI Weekly alert (citing TechCrunch), October 23, 2026: https://aiweekly.co/alerts/openai-opens-decisions-api-to-all-at-010-per-million-tokens-chases-typesafe-jev
- ai.dosa.dev daily brief, October 23, 2026: https://ai.dosa.dev/blog/today-in-ai-2026-10-23/
