FRQ Early access
Menu

Introducing Faruq-150M In development

Decisions,
not paragraphs.

A small model for the choices that keep software moving. Give Faruq a situation, a question, and your options. Get a decision, with a probability for each.

Explore the model Build with Faruq

For software, autonomous agents, and everyday workflows.

A decision, step by stepIllustrative demo
Context 
The user wants to find the cheapest roundtrip flights from New York to London departing Dec 12 and returning Dec 20.
Question

Which tool should handle this request?

Decision
  1. search_flights96.4%
  2. book_hotel2.1%
  3. currency_convert1.0%
  4. check_weather0.5%

Example inputs and probabilities. This demo does not run the model.

One pass. A structured answer.Tool selection
Architecture
150M parameter encoder
Context window
7,999 tokens per question
Output
A choice + probabilities
Speed
One pass no text generation

The model

A model built for
the moment of choice.

Which queue? Which tool? Yes or no? Faruq is a System One model: a compact encoder that scores your options in a single pass, so your software can take the next step.

Your options. Its decision.

Define the allowed choices. Faruq selects from that set, giving your code a structured result it can use directly.

No generation loop.

The encoder scores every option in one read. No generated paragraphs, extraction step, or output format to repair.

A probability for every option.

Set a threshold to accept a decision, run another check, or ask a person.

Type
Encoder
Size
150M parameters
Reads
Up to 7,999 tokens per question, situation and options included
Returns
One option, and a probability for each
Language
English; Arabic planned
Status
In development

Use cases

Small choices.
Useful consequences.

Faruq is built for the decisions inside agents and workflows: which tool, which team, what a text really says. Each one takes a single fast pass instead of a call to a large model.

Each card shows Faruq’s score on a public benchmark for that job, out of 100. See all results ↓

Give your agent a tool router.

For every tool an agent has, decide whether this request needs it, before a large model is ever called.

Route every request.

Send each message to the right intent, queue or team, and catch the ones that belong nowhere.

Turn commands into actions.

Read what someone asks a device or an app to do, and choose the right action.

Check what a contract says.

Does the agreement support a statement, contradict it, or not mention it?

Read company sentiment.

How a piece of financial news feels about each company it names.

Track stance on any topic.

Is the writer for it, against it, or neutral? For brand, policy and social listening.

Build with Faruq

One decision engine.
Two ways to build.

Use Faruq in your own software or as a step in an automation. The interface stays simple: context in, a choice out.

For developers

A decision endpoint
for your application.

Send context and questions from your code. Use the returned choices and probabilities to drive your application logic.

Explore the API preview Planned · API not yet open

For automation builders

A smarter branch
in your workflow.

Route incoming work by intent. We’re planning integrations for platforms such as n8n, Make, and Zapier.

See early-access plans Planned · Platforms to be confirmed

From incoming message to next action Workflow example

Trigger New support email "I was charged twice for my October invoice."
Faruq · decide Which team should handle this? Billing, support, sales, or account access
  • Billing94%Open a billing ticket
  • Technical support3%
  • Sales2%
  • Account access1%
Illustrative workflow and probabilities. Integrations are in planning.

Pricing planned around decisions. We’re exploring per-decision pricing and bundles. Pricing and availability are not final.

API

Context. Question. Options.
That’s the interface.

Describe the situation once. Ask one or more questions, each with its own allowed choices. Get structured answers your code can act on.

You sendpreview
You get backpreview

Proposed request and response format. The API is not open yet; details may change.

  1. State

    Whatever the decision is about: a message, a document, a conversation so far, a list of tools.

  2. Questions

    Each one is an instruction plus the options to choose from, or a plain yes or no.

  3. Answers

    For each question, the chosen option and the probability Faruq gives to every option.

Benchmarks

The evidence.
Measured, not promised.

Faruq’s own runs on public benchmarks from the Decision Index suite.

–

Tool decisions right · BFCL

For each tool on offer, does this request need it? Every tool in a case must be right.

–

Requests routed right · CLINC150

150 intents, plus the requests that fit none of them.

Where Faruq is strongest

Each benchmark’s own measure (accuracy, F1 or nDCG), out of 100. marks a benchmark whose public training set was part of Faruq’s training.

    Behind the name

    Three letters.
    One purpose.

    Faruq means the one who distinguishes. It’s a fitting name for a model learning to tell options apart.

    FRQ is Faruq without its vowels. Faruq, said fa-rook, is an Arabic name for the one who distinguishes. Its root, f-r-q, means to separate, to tell apart, to decide.

    Early access

    The next step

    Put a little more
    decision into your workflow.

    Faruq is still in development. Early-access requests will open here when signup is available.

    In development

    Follow Faruq’s progress.

    Early-access signup isn’t open yet. In the meantime, read about the model and the current evaluation results.

    About the model Explore the benchmarks