A model for each kind of decision.
Pick a model by what you send: a symbol, a piece of text, or a request and the tools to carry it out. Every release is a stable version you can pin; new updates are evaluated in a preview first.
Ora Fin
ora-fin
Send a stock or ETF symbol (or your own numbers in state.context) and named yes_no questions. Forecast questions come back as calibrated probabilities from a model trained on verified US stock and ETF history; "what is true now" questions and rules in the rule language (C > sma(C, 50)) are evaluated exactly on the latest daily bars. Text or other records in state are answered by the text engine, so one call covers both.
state.symbol or state.context, plus yes_no questions (choice and score for text) to POST /v1/systemone
probabilities and exact rule answers, with rule, basis and the version that answered
Model decision about 2 to 60 ms; a round trip through the API is typically under 200 ms
$0.042 per million input tokens, output free
The state plus every question, plus the fixed per-request overhead. · Pricing
| Version | Channel | Released | What changed |
|---|---|---|---|
ora-fin-0.1.1alias ora-fin |
Stable | 2026-10-29 | The stable financial model: forecasts for daily-bar and opening-gap questions, the rule language, and ask-by-symbol context for about 2,000 liquid US stocks and ETFs. |
QuickstartAsk by symbolRule languageExamplesGuidesCookbooksWhat's new
Ora Text
ora-text
Send a message, document or any JSON record as state and typed questions: yes_no returns the probability of yes, choice picks one of your options with a probability for each, score rates on your ordered levels. Useful for routing tickets, filtering headlines and checking claims.
text or any JSON record in state, plus yes_no, choice and score questions to POST /v1/systemone
probabilities, picks and scores with confidence
Model decision about 0.2 to 1.5 s on CPU, depending on the number of options and levels
$0.042 per million input tokens, output free
The state plus every question, plus the fixed per-request overhead. · Pricing
| Version | Channel | Released | What changed |
|---|---|---|---|
ora-text-0.2.0alias ora-text |
Stable | 2026-10-21 | The text model: a zero-shot entailment encoder that answers yes_no, choice and score about text and records. |
Ora Automation Coming soon
ora-automation
Send a query in plain English and your own tool definitions (JSON function schemas: name, description, parameters). The model picks the tools that fit and fills in their arguments, so your code runs them, and returns an empty list when no tool fits. It is built for every vertical, not just finance: smart homes, phones and assistants, cars, CRM and sales, travel booking, e-commerce, support desks, office tools and trading. A small, fast model for agents, apps and devices that need tool calling without a large language model.
query (text) plus tools: [{name, description, parameters (JSON Schema)}] to POST /v1/automate
calls: [{name, arguments}], in the order to run them
Published when it goes live
$0.042 per million input tokens, output free Introductory
Introductory: the query and your tool definitions count as input tokens, plus the fixed per-request overhead; returned calls are free.
Applies when it goes live.
| Version | Channel | Released | What changed |
|---|---|---|---|
| The first version is in internal evaluation. It is listed here when it is released. | |||
Stable, preview and aliases
ora-fin-0.1.1
A released version never changes behaviour. Pin it in production and log the version each response returns.
<family>-X.Y.Z-preview
New updates land in a preview first. Previews are internal while they are evaluated; one is released to everyone, under the stable name it was given, once it passes.
ora-fin, ora-text
Always the latest stable of a family: use one to move with each release. Older names such as ora-fin1 keep working.