Simon Willison Releases llm-openai-decisions Plugin for OpenAI's Decisions API
Simon Willison released version 0.1a0 of llm-openai-decisions, an alpha plugin for his llm CLI tool that adds support for OpenAI's newly announced Decisions API. Using GPT-6 Astra, Willison generated the plugin code based on OpenAI's documentation and his prior llm-typesafe plugin. Dedicated decision APIs represent a shift toward low-cost, high-speed classification and routing layers for AI agents that bill only for input tokens rather than text generation. This plugin enables CLI users and developers to experiment with structured multimodal decision-making directly from their terminal. OpenAI's gpt-6-luna decision model supports image inputs in addition to text and charges 10 cents per million input tokens with no output charges. The API supports three structured question types—yes/no predicates, choices, and numeric scores—returning JSON responses with calculated probabilities.
## BACKGROUND
OpenAI's Decisions API is a specialized endpoint designed for bounded classification, routing, and system evaluation tasks rather than open-ended conversational text generation. Simon Willison's llm is a popular open-source command-line utility and Python library built for interacting with LLM providers through a modular plugin system.