Models / Madmon
Embeddings

Madmon

Multilingual text embeddings specialized for Arabic dialects, Berber languages, and other underrepresented language families. Designed with deep focus on the languages frontier models ignore.

Explore recorded examples
Open Playground
Vector Size
768 dimensions
Focus
Underrepresented languages
Best For
Search and retrieval
Input
Multilingual text
Access
Sawalni API

Meaning map

A query crosses the language line.

Search in Darija

بغيت نحجز قطار للرباط

Ranked by semantic similarity

01

How do I book a train to Rabat?

0.620
02

Where can I buy a train ticket to Rabat?

0.521
03

أفضل وصفة للشاي المغربي

0.331
API capture · madmon-beta Explore the live API ↗
Read all 3 recorded examples
01

Travel query

text

Inputبغيت نحجز قطار للرباط

  1. How do I book a train to Rabat? 0.620
  2. Where can I buy a train ticket to Rabat? 0.521
  3. أفضل وصفة للشاي المغربي 0.331
02

Housing query

text

Inputكنقلب على دار للكراء فكازا

  1. Find a house to rent in Casablanca 0.516
  2. How to book a train to Tangier 0.321
  3. طريقة طهي الكسكس 0.211
03

Culture query

text

Inputبغيت نعرف تاريخ مدينة فاس

  1. A short history of Fez 0.571
  2. وصفة الحريرة المغربية 0.276
  3. Book a room in Agadir 0.220

Build with madmon

Where it fits

See the full stack →
Batbot

Retrieve source passages for grounded answers.

Search across scripts

Find passages by meaning alongside word and keyword search.

Purpose

Built for more languages

Madmon creates useful semantic representations for multilingual search, retrieval, classification, and analysis in complex linguistic environments.

It is designed with a focus on:

  • Arabic dialects — Darija, Egyptian, Gulf, Tunisian, Levantine, and more
  • Berber languages — Tamazight, Kabyle, Tachelhit in Latin, Arabic, and Tifinagh scripts
  • Other low-resource families — African, Southeast Asian, and indigenous languages

Semantic Search

Search across multilingual corpora. A Darija query retrieves relevant MSA, French, or English documents through shared semantic space.

Content Classification

Classify Arabic dialect content, social media posts, and user-generated text with high-quality representations.

Cross-Lingual Retrieval

Match content across scripts and languages. Arabizi, Arabic-script, and Latin-script content mapped to the same embedding space.

Clustering & Analytics

Group social media content by dialect, topic, or sentiment. Discover patterns in multilingual datasets.

Resources

Research Page

Try Madmon

Generate 768-dimensional semantic representations for multilingual search and content workflows.

Open Playground