reference
Models
The live catalogue is also available as JSON from GET /v1/models, which needs no authentication.
| model | alias | context | input | output | status |
|---|---|---|---|---|---|
decider-1 | sd-1 | 4,096 | $0.04 / 1M | $0.00 / 1M | Available |
restyler-1 | tr-1 | 4,096 | $0.04 / 1M | $0.16 / 1M | Available |
fanout-1 | qf-1 | 4,096 | $0.04 / 1M | $0.16 / 1M | Available |
Decider 1
Give it a piece of state — a support conversation, an alert, an invoice, an agent trace — and a set of typed questions about it. Every question comes back as a probability distribution: yes or no, a choice between labels, or a score on a rubric. Twenty questions over one state are answered in a single pass, so asking more costs almost nothing. It speaks the System One schema, so the typesafe-sdk works against it by changing its base URL.
Benchmarks and limits · Try it · task endpoint /v1/systemone
Restyler 1
Takes the response from any model, including Claude, Gemini, GPT and your own fine-tunes, and rewrites it so it reads the way a person writes. It is trained to hold on to facts, meaning, names, numbers, links and formatting — see the measured retention rates below. Built to sit at the end of a generation pipeline: cleaning up drafts, tightening documentation, and making machine-assembled copy readable.
Benchmarks and limits · Try it · task endpoint /v1/restyle
Fanout 1
Give it a question someone would put to an AI assistant, and it returns the web searches that assistant is likely to run before it answers. Built from searches observed directly from ChatGPT, Claude and Gemini rather than from queries invented by prompting a larger model. If you care whether your pages get cited in AI answers, those searches are what decide it.
Benchmarks and limits · Try it · task endpoint /v1/fanout
Short aliases
Every model also answers to a short alias — sd-1 for Decider 1, tr-1 for Restyler 1, qf-1 for Fanout 1 — anywhere a model is named: the model field of every endpoint, and API key model restrictions. An alias names exactly the same model as its full name and never moves to a different one. Responses always report the full name, so logs and dashboards stay consistent whichever form you send. The longer names the models launched under (state-decider-1, text-restyler-1, query-fanout-1) also still work.
curl https://meragpt.com/v1/fanout -H "Authorization: Bearer $MERAGPT_API_KEY" -H "Content-Type: application/json" -d '{"model": "qf-1", "question": "What is the best CRM for a small team?"}'OpenRouter
Each model has a namespaced id reserved for a future OpenRouter listing, for example meragpt/text-restyler-1. The direct API accepts both forms, so code written against the namespaced id keeps working if you point it here.