Simon Willison builds llm-openai-decisions plugin for OpenAI API

OpenAI's new Decisions API arrives after it was trailed at last week's DevDay, and it lands in a space already occupied by Jev. The two share more than a category: both charge for input only, and both expose the same three question types—yes/no, choices, or scores. That overlap makes the pricing difference the most visible point of comparison, with OpenAI at 10 cents per million input tokens and Jev at 4.2 cents per million.

The release also shows how quickly tooling can follow an API launch. Simon Willison already maintained an llm-typesafe plugin for Jev, and he had GPT-6 Astra read the new OpenAI API documentation to build a parallel llm-openai-decisions plugin. The two plugins are conceptually similar, which suggests the effort to support a second provider was modest.

The one capability gap the body notes is model support: unlike Jev, the gpt-6-luna decision model accepts image input in addition to text. Whether that feature, or the input-only pricing, ends up deciding which API teams reach for is likely to depend on the mix of text and image questions they actually run, and on how the two providers' rates evolve.