OpenAI opens Decisions API to all developers, adding image input in challenge to TypeSafe's Jev
What happened? On 6 October 2026 OpenAI made its Decisions API available to all developers in public beta, one week after previewing it to selected customers at DevDay on 29 September 2026. The API is a dedicated /v1/decisions endpoint powered by GPT-6 Luna, currently the only supported model. Instead of generating text, it evaluates an input against developer-defined questions and returns typed answers in three forms: predicates (the probability that a statement is true), choices (one option from a fixed set, with confidence scores) and scores (a probability-weighted rating against ordered levels). OpenAI says it makes decisions up to 10x faster than calling GPT-6 Luna through the Responses API, and pitches it for classifying content, routing requests and picking the right model, tool or next action for an agent. It accepts both text and images, with images passed inline as base64. Pricing is $0.10 per million input tokens, with no charges for output tokens or cache reads and writes. OpenAI expects general availability in the coming weeks.
The competitive angle: the launch lands squarely on the category that TypeSafe AI opened on 15 September 2026 with Jev, its "System One" decision model, launched alongside a $40M seed round led by DCVC. Image input is a point of difference: Jev does not yet support images, and indie developer Tony Dinh reacted to the launch on X with "They even support image input. RIP Jev". On price, TypeSafe lists Jev at $0.042 per million input tokens, under half OpenAI's rate. An early informal comparison posted by a developer on OpenAI's community forum, based on a few hundred test questions, found Jev faster (about 0.3s against 1.6s median per turn), cheaper and more accurate on nuanced judgement calls, with the two roughly level on simple factual checks. OpenAI joins a fast-growing field: Liquid AI, AWS, Perplexity and Cloudflare all released decision models or APIs between 29 September and 1 October 2026.
Why it matters: decision models are emerging as a cheap, fast layer between hand-written rules and full LLMs for automation and agents. OpenAI's distribution and multimodal support put pressure on the start-up that defined the category within three weeks of its launch, while TypeSafe still leads on published price and, in early developer tests, on speed and accuracy.
Read more: OpenAI Developer Community · OpenAI API docs · Tony Dinh on X · TypeSafe AI