JEV / SYSTEM ONE
Uses the Jev decision playground
Live OpenRouter requests with text, JSON objects and arrays pass the existing Jev validators for Choice, Score and Noul. Send model, state and a questions object to /v1/systemone. Read answers in data.result.answers. Image attachments use the same shared image interface.
SPAN / NOUL
How it differs from Span
Span-01 and Span-01 Lite accept only Noul behavior questions and require text or a specific input/output conversation object. General decision models also accept Choice and Score questions with text, JSON objects or arrays as state. Span-02 follows this broader decision contract and supports images in the shared decision playground.
Three structured answer types
Choice returns a named option and its probability distribution. Score returns a value on an ordered scale with probabilities. Noul returns the probability of a yes/no condition. Multiple questions share the same state.
choicescorenoul
OpenRouter lists a 1,050,000-token provider context and up to 200 questions. This platform accepts up to 8 questions, 2–255 Choice options and 2–10 Score levels. Text plus questions must fit 32 KiB. Add up to 4 PNG, JPEG or WebP images: 4 MiB and 16 MP each, 8 MiB total, and 13 MiB for the complete request. These platform limits apply even when the provider’s context is larger.
This integration uses OpenRouter’s decisions endpoint and model openai/gpt-6-luna-decisions. OpenAI’s native /v1/decisions uses input, question/answer arrays and predicate/probability fields, so its request format differs from the Jev contract used here. A refusal, missing answer or invalid response fails the request and releases its credit reservation.
Input and output credits
OpenRouter lists $0.10 per million input tokens and $0 for output. This platform charges 1500 credits ($0.15) per million actual input tokens, including questions and processed images, rounded up per successful request with a minimum of 1 credit ($0.0001). Web, API and evaluation share this rate. Output tokens add no charge; failed calls release the reservation.
max(1, ceil(input_tokens × 1500 / 1,000,000))
A successful request with 1,000 input tokens costs 2 credit; 5,000 input tokens cost 8 credits. Output tokens add no charge. Failed calls release the credit reservation.
Verified October 7, 2026 through OpenRouter: text, object and array state; all three answer types; 8 mixed questions; 255 Choice options accepted and 256 rejected; 10 Score levels; Noul without criteria; a long-state tail marker; contrasting red/blue images and four-image requests. Actual usage fits the reserved budget in these probes. The refusal probe returned a normal answer; refusal handling is covered by simulated route tests. These checks verify the API contract, not model accuracy or a latency SLA.