Open the interactive map โ ยท search, filter by dialect or license, share a filtered link.
Arabic AI Atlas
The Arabic AI ecosystem as a map, a list, and a skill your agent can install.
Install into your agent
claude plugin marketplace add h9-tec/arabic-ai-atlas
claude plugin install arabic-ai-atlas@arabic-ai-atlas
Requires uv on PATH.
Then ask Claude: which open Arabic TTS runs on a phone? It answers from this atlas via the recommend MCP tool.
Add this to your .mcp.json:
{"mcpServers": {"arabic-ai-atlas": {"command": "uv", "args": ["run", "--frozen", "--no-dev", "--directory", "/path/to/arabic-ai-atlas", "python", "mcp/server.py"]}}}
How this repo works
- YAML in
data/is the only hand-edited source. - CI validates every PR against
data/schema.json. - A nightly job pulls Hugging Face downloads and regenerates the map, README,
dist/atlas.jsonanddist/llms.txt. - Wanted gaps (
data/wanted.yaml) are re-evaluated on every build, andbase_modellinks build the family tree and the MCPlineagetool. - The repo is itself a Claude Code plugin: 5 skills + an MCP server that reads the atlas offline.
Entries by country
| Country | Entries | LLM | ASR | TTS | OCR | Embedding | Datasets | Tools | Benchmarks | Papers | Orgs |
|---|---|---|---|---|---|---|---|---|---|---|---|
| ๐ International | 2200 | 210 | 119 | 58 | 32 | 30 | 710 | 238 | 75 | 671 | 23 |
| ๐ธ๐ฆ Saudi Arabia | 355 | 46 | 11 | 8 | 9 | 28 | 163 | 19 | 25 | 11 | 35 |
| ๐ฆ๐ช United Arab Emirates | 184 | 49 | 5 | 6 | 0 | 0 | 43 | 12 | 31 | 17 | 21 |
| ๐ช๐ฌ Egypt | 175 | 33 | 22 | 6 | 5 | 7 | 73 | 5 | 3 | 2 | 18 |
| ๐ถ๐ฆ Qatar | 128 | 9 | 3 | 0 | 2 | 0 | 63 | 10 | 17 | 19 | 5 |
| ๐ฒ๐ฆ Morocco | 71 | 13 | 5 | 2 | 2 | 1 | 29 | 4 | 7 | 3 | 5 |
| ๐ฉ๐ฟ Algeria | 42 | 3 | 1 | 3 | 0 | 0 | 18 | 15 | 1 | 0 | 1 |
| ๐ต๐ธ Palestine | 30 | 3 | 0 | 0 | 0 | 0 | 17 | 6 | 1 | 1 | 2 |
| ๐ฏ๐ด Jordan | 25 | 1 | 1 | 0 | 1 | 0 | 15 | 2 | 2 | 1 | 2 |
| ๐ฑ๐ง Lebanon | 23 | 14 | 0 | 0 | 0 | 0 | 6 | 1 | 0 | 1 | 1 |
| ๐น๐ณ Tunisia | 22 | 2 | 1 | 1 | 0 | 0 | 11 | 0 | 1 | 3 | 3 |
| ๐พ๐ช Yemen | 9 | 1 | 2 | 0 | 0 | 0 | 4 | 0 | 1 | 1 | 0 |
| ๐ฎ๐ถ Iraq | 8 | 2 | 0 | 0 | 0 | 0 | 4 | 0 | 1 | 0 | 1 |
| ๐ด๐ฒ Oman | 8 | 2 | 0 | 0 | 0 | 0 | 2 | 1 | 0 | 0 | 3 |
| ๐ธ๐ฉ Sudan | 7 | 0 | 1 | 0 | 0 | 0 | 3 | 0 | 0 | 1 | 2 |
| ๐ง๐ญ Bahrain | 5 | 1 | 2 | 0 | 0 | 0 | 1 | 0 | 0 | 0 | 1 |
| ๐ฐ๐ผ Kuwait | 4 | 1 | 0 | 0 | 0 | 0 | 1 | 0 | 0 | 1 | 1 |
| ๐ฑ๐พ Libya | 3 | 1 | 1 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 1 |
| ๐ธ๐พ Syria | 3 | 2 | 0 | 1 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
Grid view
The same atlas as a country-by-type grid, with the most downloaded entries named in each cell.
Contents
- ๐ง Large Language Models
- ๐๏ธ Speech Recognition
- ๐ Text-to-Speech
- ๐ OCR
- ๐ค Embeddings
- ๐ Datasets
- ๐ง Tools
- ๐ Benchmarks
- ๐ Papers
- ๐ข Organizations
- ๐งฉ Arabic Agent Skills
- ๐ฏ Most Wanted
- ๐ณ Family tree
- ๐ค Contributing
๐ง Large Language Models
Showing 20 of 393 ยท see all 393 on the interactive map ยท full table
๐๏ธ Speech Recognition
Showing 20 of 174 ยท see all 174 on the interactive map ยท full table
๐ Text-to-Speech
Showing 20 of 85 ยท see all 85 on the interactive map ยท full table
๐ OCR
Showing 20 of 51 ยท see all 51 on the interactive map ยท full table
๐ค Embeddings
Showing 20 of 66 ยท see all 66 on the interactive map ยท full table
๐ Datasets
Showing 20 of 1163 ยท see all 1163 on the interactive map ยท full table
๐ง Tools
Showing 20 of 313 ยท see all 313 on the interactive map ยท full table
๐ Benchmarks
Showing 20 of 165 ยท see all 165 on the interactive map ยท full table
๐ Papers
Showing 20 of 732 ยท see all 732 on the interactive map ยท full table
๐ข Organizations
Showing 20 of 125 ยท see all 125 on the interactive map ยท full table
๐งฉ Arabic Agent Skills
Skills shipped in this repo:
| Skill | What it does | Install |
|---|---|---|
| arabic-ai-advisor | Use when choosing an Arabic LLM, ASR, TTS, OCR, embedding model, dataset or tool โ recommends ranked options from the Arabic AI Atlas with licenses, dialect coverage and on-device fit. | skills/arabic-ai-advisor |
| arabic-dialect-prompts | Use when writing system prompts or few-shot examples for an Arabic-speaking agent in a specific dialect, or when a model answers in formal Arabic (MSA) instead of the requested dialect. | skills/arabic-dialect-prompts |
| arabic-token-cost | Use when estimating or comparing how many tokens Arabic text costs across LLM tokenizers, or when Arabic prompts seem expensive, truncated or slow compared with English. | skills/arabic-token-cost |
| rtl-bidi-lint | Use when reviewing or generating Arabic text, UI strings, HTML or Markdown that mixes RTL with LTR content, to catch digit mixing, unbalanced bidi isolates, stray LTR marks and hardcoded dir="ltr". | skills/rtl-bidi-lint |
| tashkeel-check | Use when Arabic text may need diacritics (tashkeel) added or stripped, for TTS input, learner material, Quran or poetry, or when diacritization looks partial or inconsistent. | skills/tashkeel-check |
Other Arabic agent skills in the wild:
Showing 20 of 35 ยท see all 35 on the interactive map ยท full table
๐ฏ Most Wanted
Gaps nobody has filled yet. Add an entry that matches a rule and it moves to Recently filled. Rules live in data/wanted.yaml.
| Gap | Why | Rule |
|---|---|---|
| Open Arabic TTS with a Kuwaiti or Qatari voice | No openly licensed text-to-speech model with a Kuwaiti or Qatari voice is listed. | type=tts ยท country=KW,QA ยท license=open |
| Open Moroccan Arabic OCR dataset | The only Moroccan OCR dataset listed has no clear license, so no openly licensed one is available. | type=dataset ยท country=MA ยท license=open ยท tasks=ocr |
| Sudanese Arabic speech dataset for ASR | A Sudanese ASR model exists, but no Sudanese speech dataset to train or test one is listed. | type=dataset ยท country=SD ยท tasks=asr |
| Anything from Mauritania | Mauritania has no listed model, dataset, or tool of any kind. | type=llm,asr,tts,ocr,embedding,dataset,tool,benchmark ยท country=MR |
| Open Yemeni speech dataset | No openly licensed Yemeni dataset for speech recognition or synthesis is listed. | type=dataset ยท country=YE ยท license=open ยท tasks=asr,speech,tts |
| Embedding model for Sudanese, Yemeni or Iraqi dialects | No Arabic embedding model targets the Sudanese, Yemeni, or Iraqi dialects. | type=embedding ยท dialects=sudanese,yemeni,iraqi |
| Open-licensed Iraqi Arabic TTS | The only TTS covering Iraqi has no clear license, so no openly licensed one is available. | type=tts ยท dialects=iraqi ยท license=open |
| Libyan Arabic corpus | No dataset from or about Libya is listed. | type=dataset ยท country=LY |
| Open Syrian-dialect dataset | No openly licensed dataset of Syrian Levantine Arabic is listed. | type=dataset ยท country=SY ยท dialects=lev ยท license=open |
| Palestinian Arabic speech recognition | No speech recognition model from Palestine is listed. | type=asr ยท country=PS |
| Open Omani dialect dataset | Omani datasets are listed, but none has a clear open license. | type=dataset ยท country=OM ยท license=open |
| Open Kuwaiti Arabic dataset | The only Kuwaiti dataset listed has no clear license. | type=dataset ยท country=KW ยท license=open |
| Open Bahraini Arabic dataset | The only Bahraini dataset listed is non-commercial, so no openly licensed one is available. | type=dataset ยท country=BH ยท license=open |
| Resources from the Horn of Africa and Comoros | Nothing from Comoros, Djibouti, or Somalia is listed. | country=KM,DJ,SO |
| Open Arabic handwriting benchmark | No openly licensed benchmark covers Arabic handwriting recognition. | type=benchmark ยท license=open ยท tasks=handwriting,htr |
๐ณ Family tree
385 models with known lineage ยท explore the tree ยท add a base model with the base_model field (CONTRIBUTING)
| Base family | Models | Most downloaded |
|---|---|---|
| Qwen ยท ูููู | 65 | qwen3-asr-arabic-uae (62K โฌ) |
| BERT ยท ุจูุฑุช | 58 | AraBERTv02 (541K โฌ) |
| Whisper ยท ููุณุจุฑ | 56 | whisper-large-v3-turbo (5.8M โฌ) |
| Gemma ยท ุฌูู ุง | 24 | gemma4 e4b claims comparison (248K โฌ) |
| From scratch ยท ู ู ุงูุตูุฑ | 19 | ALLaM (92K โฌ) |
| Llama ยท ูุงู ุง | 15 | Llama 3.3 (209K โฌ) |
| Cohere | 9 | cohere-transcribe-arabic-07-2026 (161K โฌ) |
| T5 ยท ุชู 5 | 9 | AraT5v2-base-1024 (1K โฌ) |
| XTTS | 7 | XTTS-v2 (6.7M โฌ) |
| BGE | 4 | arabic english bge m3 (1K โฌ) |
| F5-TTS | 4 | f5tts-algerian-darja (12K โฌ) |
| XLS-R ยท ุฅูุณ ุฅู ุฅุณ-ุขุฑ | 4 | egyptian-arabic-wav2vec2-xlsr-53 (5K โฌ) |
| NLLB | 3 | darija to english 2 (25 โฌ) |
| SeamlessM4T | 2 | SeamlessM4T v2 (310K โฌ) |
| wav2vec ยท ูุงู ุชู ููู | 2 | wav2vec2 quran phonetics (114K โฌ) |
| BLOOM ยท ุจููู | 1 | darija text generation (21 โฌ) |
| E5 | 1 | algerianME5 (82 โฌ) |
| Falcon ยท ูุงูููู | 1 | Falcon Arabic (11K โฌ) |
| Mistral ยท ู ูุณุชุฑุงู | 1 | Mistral 7B v0.1 arabic (21 โฌ) |
| MMS ยท ุฅู ุฅู ุฅุณ | 1 | mms 300m arabic dialect identifier (1K โฌ) |
| Other ยท ุฃุฎุฑู | 99 | Multilingual Chatterbox (1.7M โฌ) |
๐ค Contributing
- Edit the YAML in
data/. - Run
uv run python scripts/build.py build(it uses the committed Hugging Face cache; the nightly job refreshes metrics). - Open a PR.
Never hand-edit README.md, docs/tables/, assets/, dist/ or data/.cache/: the build and the nightly job write them, and CI rejects drift.
๐ License
Code: MIT. Data: CC BY 4.0.
๐ Sister project
Awesome Arabic NLP, the curated reading list this atlas grew alongside.
Generated 2026-10-27 from 3302 entries. Do not edit README.md by hand.