HiNow LogoHiNow

Pantala

Model Information

Display Name: Pantala

API Model ID: hinow/pantala

Category: Text To Decision

Description: Pantala is the HiNow decision model: one stable name for fast, structured decisions inside your software. It does not write text. You give it a state and ask typed questions, and it answers with typed decisions and calibrated probabilities — in a fraction of a second, at a fraction of the cost of an LLM call. **Endpoint:** `POST /v1/decisions` (not chat completions) **How it works:** - `state` — the content to evaluate: a string, a JSON object or an array - `questions` — one or more named questions, all answered in a single call **Question types:** - **noul** — a yes/no question, answered as a probability from 0 to 1. The shortest form is just a string: `"urgent": "Does this convey urgency?"` - **choice** — pick one option from a set you define (2 to 255 options). Returns the choice, a probability for every option and a confidence - **score** — rate the state on an ordered scale you describe (2 to 10 levels). Returns a score, a probability for every level and a confidence **Sensible defaults:** - `type` is optional — it is inferred from `criteria`: none means yes/no, an array means score, an object means choice - A choice accepts a plain list of option names when the names explain themselves - Anything that cannot be inferred is rejected before it is billed, with a message that names the question and the fix **Key Features:** - Typed, schema-safe answers your code can use directly — nothing to parse - Calibrated probabilities, so you can set thresholds and escalate uncertain cases - As many questions per request as fit the 32,000-token context, evaluated in parallel - Structured instructions and criteria (JSON objects and arrays) when a question needs a rubric - Only input tokens are billed — answers are free - Understands natural language in many languages, including Portuguese **Best For:** - Model routing and intent classification - LLM guardrails: jailbreak, policy and tool-call checks - Support ticket triage and lead scoring - Search reranking and relevance scoring - Classifying and extracting features from large datasets **Note:** text input only. There are no sampling parameters — answers are deterministic by design.

Context Window: 32,000 tokens

How to Use This Model

To use Pantala via the HInow.ai API, use the model ID: hinow/pantala

API Request Example (Chat/Text)


POST https://api.hinow.ai/v1/chat/completions
Authorization: Bearer YOUR_API_KEY
Content-Type: application/json

{
  "model": "hinow/pantala",
  "messages": [
    {"role": "user", "content": "Your message here"}
  ]
}

Pricing

  • input: $0.126

Available Parameters

  • state: The content to evaluate: a plain string, or a JSON object or array of related context. Cannot be empty; numbers and booleans are not accepted
  • questions: Named questions, all answered in one call and evaluated in parallel. You choose each name and read the answer back under it; the name is never shown to the model. No fixed limit: state and questions share the 32,000-token context
  • questions.*: One question. Shortest form: a plain string, which is a yes/no (noul) question. Full form: an object with instructions, and optionally type and criteria
  • questions.*.type: choice = pick one option; score = rate on an ordered scale; noul = yes/no as a probability. Optional: when omitted it is inferred — no criteria = noul, {true,false} object = noul, any other object = choice, array = score (Options: choice, score, noul)
  • questions.*.instructions: The question itself, in natural language. Refer to parts of a structured state by path, e.g. `ticket.messages[0].text`. May be a JSON object or array when the question has several labeled parts
  • questions.*.criteria: choice (required): object {option: meaning}, 2 to 255 options; the meaning may be null when the name says it all, and a plain array of option names is accepted. score (required): ordered array of level descriptions, lowest first, 2 to 10 levels. noul (optional): object with both "true" and "false" descriptions. Descriptions may be strings, objects or arrays

Quick Reference

To use this model, set: "model": "hinow/pantala"

Featured: No

Documentation: https://hinow.ai/models/hinow/pantala

API Endpoint: https://api.hinow.ai/v1

Back to Models
Pantala

Pantala

hinow/pantala

$0.126
input

About

Pantala is the HiNow decision model: one stable name for fast, structured decisions inside your software. It does not write text. You give it a state and ask typed questions, and it answers with typed decisions and calibrated probabilities — in a fraction of a second, at a fraction of the cost of an LLM call.

Endpoint: POST /v1/decisions (not chat completions)

How it works:

  • state — the content to evaluate: a string, a JSON object or an array
  • questions — one or more named questions, all answered in a single call

Question types:

  • noul — a yes/no question, answered as a probability from 0 to 1. The shortest form is just a string: "urgent": "Does this convey urgency?"
  • choice — pick one option from a set you define (2 to 255 options). Returns the choice, a probability for every option and a confidence
  • score — rate the state on an ordered scale you describe (2 to 10 levels). Returns a score, a probability for every level and a confidence

Sensible defaults:

  • type is optional — it is inferred from criteria: none means yes/no, an array means score, an object means choice
  • A choice accepts a plain list of option names when the names explain themselves
  • Anything that cannot be inferred is rejected before it is billed, with a message that names the question and the fix

Key Features:

  • Typed, schema-safe answers your code can use directly — nothing to parse
  • Calibrated probabilities, so you can set thresholds and escalate uncertain cases
  • As many questions per request as fit the 32,000-token context, evaluated in parallel
  • Structured instructions and criteria (JSON objects and arrays) when a question needs a rubric
  • Only input tokens are billed — answers are free
  • Understands natural language in many languages, including Portuguese

Best For:

  • Model routing and intent classification
  • LLM guardrails: jailbreak, policy and tool-call checks
  • Support ticket triage and lead scoring
  • Search reranking and relevance scoring
  • Classifying and extracting features from large datasets

Note: text input only. There are no sampling parameters — answers are deterministic by design.

Capabilities

Text To Decision
Context32K tokens

Parameters

state

The content to evaluate: a plain string, or a JSON object or array of related context. Cannot be empty; numbers and booleans are not accepted

questions

Named questions, all answered in one call and evaluated in parallel. You choose each name and read the answer back under it; the name is never shown to the model. No fixed limit: state and questions share the 32,000-token context

questions.*

One question. Shortest form: a plain string, which is a yes/no (noul) question. Full form: an object with instructions, and optionally type and criteria

questions.*.type

choice = pick one option; score = rate on an ordered scale; noul = yes/no as a probability. Optional: when omitted it is inferred — no criteria = noul, {true,false} object = noul, any other object = choice, array = score

choicescorenoul
questions.*.instructions

The question itself, in natural language. Refer to parts of a structured state by path, e.g. `ticket.messages[0].text`. May be a JSON object or array when the question has several labeled parts

questions.*.criteria

choice (required): object {option: meaning}, 2 to 255 options; the meaning may be null when the name says it all, and a plain array of option names is accepted. score (required): ordered array of level descriptions, lowest first, 2 to 10 levels. noul (optional): object with both "true" and "false" descriptions. Descriptions may be strings, objects or arrays

Code Examples

curl -X POST https://api.hinow.ai/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $HINOW_API_KEY" \
  -d '{
    "model": "hinow/pantala",
    "messages": [
      {"role": "user", "content": "Your input here"}
    ],
    "parameters": {
      "state": "",
      "questions": "",
      "questions.*": "",
      "questions.*.type": "choice",
      "questions.*.instructions": "",
      "questions.*.criteria": ""
    }
  }'