This page requires JavaScript to display.

OpenAI's Decisions API

AI & tech · Heat 1.5 / 10 · 3 shows · 9m airtime

At Dev Day on September 30, 2026, OpenAI introduced the Decisions API, a preview tool using a version of Luna that it says is 10 times faster than the Responses API and supports visual inputs. Reporting since has established that it is a smaller model without reasoning, built for parallel inference and weaker on long-horizon tasks, and that it was built on existing Luna weights rather than a newly trained model. It is now generally available at $0.10 per 1M input tokens with no output charges. What the speed, visual support and pricing mean in practice remains the open question.

How 3 podcasts covered it

  1. The AI Daily Brief, 2026-09-30 (6:22 to 8:12)
    Reports the Decisions API's speed and inputs
    Covers the launch of OpenAI's Decisions API at Dev Day, a preview tool that mimics JEV's classification using a version of Luna, claimed 10 times faster than the Responses API and supporting visual inputs.
    New angle: OpenAI's Decisions API uses a version of Luna, is 10x faster than the Responses API, and supports visual inputs.
  2. Latent Space, 2026-09-30 (1:35 to 2:24)
    Contrasts the Decisions API with computer use
    Ari Weinstein contrasts the Decisions API with computer-use models: it is a smaller model without reasoning, built for parallel inference, and weaker on long-horizon tasks. He frames combining the two approaches as an open research area.
    New angle: The Decisions API is a smaller model without reasoning, built for parallel inference, and weaker on long-horizon tasks.
  3. Latent Space, 2026-09-30 (23:35 to 25:13)
    Asks how the Decisions API was built
    Asks how fast OpenAI answered Jev and gets the origin story: the Decisions API was built on existing Luna weights rather than a new model, and internal teams wanted a faster classifier. Adds the Jev team's role and the latency-target launch plan.
    New angle: The Decisions API was built on existing Luna weights rather than a newly trained model.
  4. Tech Brew Ride Home, 2026-10-23 (8:24 to 12:37)
    Covers the Decisions API launch and pricing
    Covers the Decisions API's general availability: pricing at $0.10 per million input tokens with free output, latency down to 120-180 milliseconds, and named customers Box, Datadog, Ramp and Speak replacing LLMs. Adds benchmark placements and Honda's remarks on multimodality.
    New angle: The Decisions API is generally available at $0.10 per 1M input tokens with no output charges.
Unpacking...