📑 Table of Contents
🏢 Key Organizations
Research labs and institutions driving Arabic NLP forward.
| Organization | Focus | Key Contributions | Link |
|---|
| AUB MIND Lab | Foundational Arabic NLP models | AraBERT, AraGPT2, AraELECTRA |  |
| CAMeL Lab, NYUAD | Arabic NLP tools and models | CAMeLBERT, camel_tools |  |
| UBC-NLP | Dialectal Arabic, multimodal models | MARBERT, AraT5, NileChat, PEARL, Dallah |  |
| QCRI | Arabic LLMs, text processing | Fanar LLMs, AraDiCE, Farasa |  |
| SILMA AI | State-of-the-art Arabic LLMs | SILMA LLMs, Arabic Broad Benchmark |  |
| MBZUAI | Multimodal and speech models | AIN, ArTST, ClArTTS |  |
| ARBML | Democratizing Arabic NLP | masader, klaam, tkseem |  |
| NAMAA-Space | Arabic NLP models & dialect hub | Qari-OCR, EgypTalk-ASR, Masrawy translator, GLiNER Arabic |  |
| ITIDA / MCIT (Egypt) | Egypt's national AI, sovereign models | Karnak LLM, BelMasry, Torgoman, SIA, AcQua |  |
| Omartificial-Intelligence-Space | Arabic embedding models | GATE, Matryoshka embeddings |  |
| TII (Technology Innovation Institute) | Arabic LLM benchmarks, Falcon | Open Arabic LLM Leaderboard, Falcon LLM |  |
| SDAIA (Saudi Data & AI Authority) | Sovereign Arabic LLM, national AI strategy | ALLaM model, SADA dataset, BALSAM benchmark |  |
| HUMAIN (Saudi PIF) | Full-stack AI company | ALLaM 34B, HUMAIN Chat, AI factories |  |
| KAUST | AI research, Arabic NLP | Center of Excellence in Generative AI |  |
| KFUPM-JRCAI | Joint SDAIA-KFUPM AI research | Arabic AI text detection datasets |  |
| Prince Sultan University (RIOTU) | Arabic language models | ArabianGPT, Arabic IoT AI |  |
| SinaLab, Birzeit University | Arabic NLP tools and datasets | SinaTools, Wojood NER |  |
| G42 / Inception AI | Arabic-centric LLMs | Jais LLM family |  |
| FreedomIntelligence | Arabic LLMs and alignment | AceGPT, Arabic cultural datasets |  |
| Helsinki-NLP | Machine translation models | OPUS-MT Arabic translation models |  |
| LightOn AI | Arabic web data | ArabicWeb24 corpus (39B+ tokens) |  |
🏆 Benchmarks & Leaderboards
Standardized evaluation frameworks for Arabic language models.
| Benchmark | Description | Link |
|---|
| MTEB Arabic Leaderboard | Massive Text Embedding Benchmark for Arabic |  |
| Arabic Broad Leaderboard (ABL) | NextGen evaluation for Arabic LLMs by SILMA AI |  |
| Open Arabic LLM Leaderboard | Evaluation of Arabic LLMs across multiple benchmarks |  |
| ALUE | Arabic Language Understanding Evaluation benchmark |  |
| BALSAM | Benchmark of Arabic Language AI Systems and Models |  |
| SILMA RAGQA Benchmark | Evaluates Arabic/English LMs in Extractive QA tasks |  |
| Arabic Broad Benchmark (ABB) | Comprehensive evaluation tool for Arabic LLMs |  |
| ArabicMMLU | Multi-task language understanding from school exams |  |
| GATmath and GATLc | Benchmarks from Saudi GAT exams |  |
| ArabicRAGB | Arabic RAG Benchmark (multi-dialect) |  |
| ACVA | Arabic Cultural Value Alignment (8000+ questions, 58 areas) |  |
| AraDiCE | Benchmarks for dialectal and cultural capabilities of LLMs |  |
| Open Universal Arabic ASR Leaderboard | Multi-dialectal Arabic speech recognition benchmark |  |
| KITAB-Bench | Arabic OCR benchmark: 8,809 samples, 9 domains, 36 sub-domains (MBZUAI) |  |
🤖 State-of-the-Art Models
💬 Large Language Models (LLMs)
Arabic-centric and multilingual LLMs with strong Arabic capabilities.
| Model | Params | Developer | Key Features | Link |
|---|
| Jais | 13B, 30B | Inception AI, Cerebras | Arabic-centric, bilingual, instruction-tuned |  |
| SILMA 1.0 | 9B | SILMA AI | Top-ranked Arabic LLM built on Gemma |  |
| ALLaM | 7B | SDAIA & IBM | Saudi's sovereign model, enterprise-focused |  |
| Fanar-1-9B | 9B | QCRI | Arabic-English LLM |  |
| AceGPT | 7B | FreedomIntelligence | Top performance, culturally aligned |  |
| Atlas-Chat | 2B-27B | MBZUAI-Paris Lab | Moroccan Darija dialect |  |
| NileChat-3B | 3B | UBC-NLP | Egyptian and Moroccan dialects |  |
| Nile-Chat | 4B-12B | MBZUAI-Paris Lab | Egyptian Arabic and Arabizi scripts |  |
| AraGPT2 | 1.5B | AUB MIND Lab | GPT-2 for Arabic text generation |  |
| Command R7B Arabic | 7B | Cohere | Arabic-optimized Command R variant |  |
| SambaLingo-Arabic | 7B, 70B | SambaNova | Arabic-adapted Llama 2 |  |
| ArabianGPT | 0.1B | Prince Sultan University | GPT-2 for Arabic, RIOTU Lab |  |
| Nuha | - | Elm (Saudi) | Multi-modal Arabic-first LLM for gov services, dialect-aware |  |
| Llama 3.3 | 70B | Meta | Strong Arabic performance |  |
| Qwen 3 | 0.6B-235B | Alibaba | Multilingual with Arabic support |  |
| Gemma 3 | 1B-27B | Google | Multimodal capabilities |  |
| Cohere Command-A | 111B | Cohere | Optimized for RAG |  |
| Mistral Saba | 24B | Mistral | Commercial API |  |
| Falcon Arabic | 7B | TII (UAE) | First Arabic model in Falcon series, top of OALL |  |
| Falcon-H1-Arabic | 3B-34B | TII (UAE) | Hybrid Mamba-Transformer, 128K-256K context |  |
| ALLaM-2-7B | 7B | SDAIA & IBM | 500B+ Arabic tokens, largest Arabic training set |  |
| Mulhem | - | SDAIA | Open-source Arabic-first LLM from Saudi Arabia |  |
| Arcee-Meraj | 72B | Arcee AI | Enterprise Arabic LLM based on Qwen2-72B |  |
| Meraj-Mini | 7B | Arcee AI | Arabic-capable Meraj family (public checkpoint) |  |
| Aya-Expanse | 8B-32B | Cohere | State-of-the-art multilingual with strong Arabic |  |
| SILMA Kashif | 2B | SILMA AI | Lightweight RAG-optimized Arabic model |  |
| Ar-stablelm-2-chat | 1.6B | Stability AI | Small Arabic chat model |  |
| Yehia | 7B | Navid-AI | Based on ALLaM, Arabic instruction-tuned |  |
| AraLLaMA | 7B | Bashar Alhafni | Arabic adaptation via progressive vocabulary expansion |  |
| Shahin | 14B | malhajar | Syrian Arabic dialect, Qwen2.5-based |  |
| Labess Chat | 7B | Linagora | Tunisian Arabic, based on Jais architecture |  |
| Phi-4 | 4B | Microsoft | Multilingual with Arabic, compact & efficient |  |
| ALLaM 34B | 34B | HUMAIN (Saudi) | Most advanced Arabic LLM, 8PB training data, powers HUMAIN Chat |  |
| Karnak | 30B-70B | ITIDA (Egypt) | Egypt's national sovereign LLM, top Arabic in its class |  |
| NOOR | 10B | TII (UAE) | World's largest Arabic NLP model at launch, GPT-3 architecture |  |
| Noon | 7B | Naseej | BLOOM-based Arabic LLM, instruction-tuned |  |
| Fanar Star | 7B | QCRI (Qatar) | Trained from scratch on 1T Arabic/English tokens |  |
| JASMINE | 0.3B-6.7B | UBC-NLP | Arabic GPT for few-shot learning, 400GB training data |  |
| Kawn | - | Misraj AI (Saudi) | Arabic-first ecosystem with 15-dialect support |  |
| Kuwain | 1.5B | Misraj AI | Arabic SLM via language injection, 70% cost reduction |  |
| Mutarjim | 1.5B | Misraj AI | Arabic-English translation, rivals GPT-4o mini |  |
| GemmAr | 7B | ClusterlabAi | Gemma Arabic instruction-tuned on InstAr-500k |  |
| LlamAr | 8B | ClusterlabAi | LLaMA 3 Arabic instruction-tuned on InstAr-500k |  |
| Calme 2.2 | 72B | MaziyarPanahi | Qwen2-based, strong OALL performance |  |
🌐 Multimodal Models
| Model | Params | Developer | Key Features | Link |
|---|
| AIN | 8B | MBZUAI | Arabic-centric Large Multimodal Model |  |
| Dallah | - | UBC-NLP | Advanced multimodal LLM for Arabic |  |
| Peacock | 7B | UBC-NLP | Arabic multimodal (InstructBLIP + AraLLaMA) |  |
| Qalam | - | - | Arabic OCR/HWR multimodal LLM, 0.80% WER |  |
| Model | Developer | Key Features | Link |
|---|
| AraBERT | AUB MIND Lab | First BERT for Arabic, multiple versions |  |
| AraBERTv02 | AUB MIND Lab | Improved tokenization (135M params, 12M+ downloads) |  |
| CAMeLBERT | CAMeL Lab, NYUAD | MSA, Dialectal, and Classical Arabic |  |
| MARBERT | UBC-NLP | Focused on Dialectal Arabic and MSA |  |
| MARBERTv2 | UBC-NLP | Updated with improved dialectal coverage |  |
| AraELECTRA | AUB MIND Lab | ELECTRA for Arabic |  |
| AraT5 | UBC-NLP | T5 for Arabic summarization, translation, paraphrasing |  |
| SaudiBERT | King Saud University | Saudi dialect BERT, trained on 141M tweets (STMC) + forums |  |
| EgyBERT | faisalq | Egyptian dialect BERT, trained on 34M tweets + 44M forum sentences |  |
| SA-BERT-V1 | Omartificial-Intelligence-Space | Saudi Arabic BERT, fine-tuned MARBERTv2 |  |
🔤 Embedding Models
| Model | Developer | Key Features | Link |
|---|
| GATE-AraBert-v1 | Omartificial-Intelligence-Space | SOTA on MTEB Arabic STS |  |
| Arabic-Triplet-Matryoshka-V2 | Omartificial-Intelligence-Space | Matryoshka representation |  |
| Swan | UBC-NLP | Dialect-aware, cross-lingual |  |
| multilingual-e5-base | intfloat | Strong multilingual embeddings baseline (incl. Arabic) |  |
| asafaya/bert-base-arabic | asafaya | BERT-based Arabic embeddings |  |
| ModernBERT-Arabic | BounharAbdelaziz | ModernBERT-based sentence embeddings |  |
| DIMI-embedding | AhmedZaky1 | Matryoshka + AraBERT for NLI |  |
🎯 Task-Specific Models
| Model | Task | Key Features | Link |
|---|
| CAMeLBERT-MSA-Sentiment | Sentiment Analysis | Fine-tuned for MSA sentiment |  |
| t5-arabic-summarization | Summarization | T5 for Arabic news summarization |  |
| opus-mt-en-ar | Translation EN→AR | Helsinki-NLP (3.5M+ downloads) |  |
| opus-mt-ar-en | Translation AR→EN | Helsinki-NLP (12.4M+ downloads) |  |
| arabic-gec-v1 | Grammar Correction | Gemma-3-1b for Arabic GEC |  |
| Arabic-Text-Correction | Text Correction | AraT5-based text correction |  |
| arat5-dialects-translation | Dialect→MSA | AraT5 dialect translation |  |
| Masrawy Translator | EN→Egyptian Arabic | 150K+ rows, 10M+ tokens, Egyptian dialect |  |
| GLiNER Arabic | Named Entity Recognition | Plug-and-play Arabic NER (NAMAA-Space) |  |
🎙️ Audio Models
🗣️ Speech Recognition (ASR)
| Model | Key Features | Link |
|---|
| openai/whisper-large-v3 | Supports Arabic among many languages |  |
| MasriSwitch-Gemma3n | Egyptian Arabic code-switching transcription |  |
| wav2vec2-large-xlsr-53-arabic | Fine-tuned on Common Voice & Arabic Speech Corpus |  |
| artst_asr_v3 | ArTST for ASR on MGB2 (best for MSA) |  |
| EgypTalk-ASR-v2 | High-performance ASR for Egyptian Arabic |  |
| whisper-large-v3-turbo | 4x faster Whisper, Arabic support, MIT license |  |
| NVIDIA FastConformer Arabic | SOTA Arabic ASR, 115M params, 760h training, CC-BY-4.0 |  |
| NVIDIA FastConformer Arabic (Diacritics) | Arabic ASR with diacritical marks, 1100h training |  |
| SeamlessM4T v2 | Meta's all-in-one ASR + translation, ~100 languages inc. Arabic |  |
| MMS-1b-all | Meta's Massively Multilingual Speech, ASR for 1100+ languages |  |
| SpeechBrain wav2vec2 Arabic | Wav2Vec2 + CTC fine-tuned on CommonVoice 14 Arabic |  |
| HuBERT-Large Arabic | HuBERT fine-tuned on 2000h Arabic, 17.68% WER |  |
| HuBERT Egyptian Arabic | HuBERT fine-tuned for Egyptian dialect on MGB-3 |  |
| Whisper Quran | Whisper fine-tuned for Quranic recitation recognition |  |
| Whisper Egyptian Arabic | Whisper-medium fine-tuned on 72h Egyptian speech |  |
| Whisper Arabic (small) | Whisper-small fine-tuned for Arabic language |  |
| KalemaTech Arabic STT | Whisper-small optimized for Arabic STT |  |
| Voxtral Mini | Mistral's speech model, 3B, Arabic support, Apache 2.0 |  |
| Klaam | Arabic ASR/TTS/classification library (MSA + Egyptian) |  |
🔊 Text-to-Speech (TTS)
| Model | Key Features | Link |
|---|
| facebook/mms-tts-ara | Facebook's Massively Multilingual Speech |  |
| speecht5_tts_clartts_ar | SpeechT5 for Classical Arabic TTS |  |
| F5-TTS-Arabic | F5-TTS with regional diversity |  |
| Arabic-F5-TTS-v2 | Improved F5-TTS Arabic fine-tune |  |
| Arabic-TTS-Spark | Spark TTS fine-tuned on 300h clean Arabic, MSA with diacritics |  |
| LLMVoX | 30M streaming TTS for any LLM, first Arabic autoregressive TTS |  |
| XTTS-v2 | Coqui's multilingual voice cloning, Arabic support, 6s cloning |  |
| Arabic MMS Speech Synthesis | VITS-based Arabic TTS from MMS (36M params) |  |
| tts-arabic-pytorch | Tacotron2 + FastPitch + HiFi-GAN for Arabic |  |
| tts_arabic (ONNX) | FastPitch + Mixer-TTS in ONNX for offline Arabic TTS |  |
| Arabic Tacotron TTS | End-to-end Arabic TTS based on Tacotron |  |
| Piper TTS | Fast local neural TTS, Arabic voices available |  |
📊 Arabic TTS Datasets & Benchmarks
| Resource | Description | Link |
|---|
| ArVoice | Multi-speaker MSA corpus, 83h, 11 voices, diacritized |  |
| ClArTTS | Classical Arabic TTS corpus by MBZUAI |  |
| Arabic Speech Corpus | Nawar Halabi's Levantine Arabic TTS corpus (CC-BY-4.0) |  |
| SawtArabi | First Arabic dialectal + code-switching TTS benchmark |  |
| Open-Source Arabic TTS Benchmark | SILMA AI's auditory assessment benchmark |  |
👁️ Vision Models
📖 Optical Character Recognition (OCR)
| Model | Key Features | Link |
|---|
| QARI-OCR v0.2 | Updated Qwen2-VL 2B for Arabic OCR; WER 0.160, CER 0.061 |  |
| Qari-OCR-0.1-VL-2B-Instruct | Qwen2 VL 2B fine-tuned for Arabic OCR |  |
| Baseer | Arabic document-to-markdown, Qwen2.5-VL-3B based (Misraj AI), WER 0.25 |  |
| Arabic-English-handwritten-OCR-v3 | Qwen2.5-VL-3B, trained on 47K handwriting samples |  |
| AtlasOCR | First Darija/Moroccan Arabic OCR, Qwen2.5-VL-3B based |  |
| DIMI-Arabic-OCR | Printed Arabic with diacritics OCR (Qwen2-VL based) |  |
| arabic-large-nougat | End-to-end structured OCR for Arabic documents |  |
| arabic-small-nougat | Smaller Nougat variant for Arabic document OCR |  |
| Tool | Key Features | Link |
|---|
| PaddleOCR | High-performance multilingual OCR with Arabic support |  |
| EasyOCR | Ready-to-use OCR with Arabic support (80+ languages) |  |
| Tesseract OCR | Open-source OCR engine with Arabic language packs |  |
📊 Arabic OCR Datasets & Benchmarks
| Resource | Description | Link |
|---|
| KITAB-Bench | 8,809 samples across 9 domains & 36 sub-domains (MBZUAI, ACL 2025) |  |
| SARD | Synthetic Arabic OCR dataset for recognition training |  |
| Misraj-DocOCR Benchmark | Arabic document OCR evaluation benchmark by Misraj AI |  |
🖼️ Image Captioning
| Model | Key Features | Link |
|---|
| blip-Arabic-flickr-8k | BLIP fine-tuned for Arabic image captioning |  |
✏️ Diacritization (Tashkeel)
Models and tools for adding harakat (حركات) to Arabic text.
Models
| Model / System | Key Features | Link |
|---|
| CATT | Character-based Tashkeel Transformer, SOTA results |  |
| Fine-Tashkeel | Fine-tuned ByT5, 40% WER reduction |  |
| Sadeed | Small language model for diacritization |  |
| Shakkala | Neural vocalization using bidirectional LSTM |  |
| Mishkal | Rule-based diacritizer with dictionary lookups |  |
Datasets
| Dataset | Description | Link |
|---|
| Tashkeela | Arabic diacritization corpus |  |
| arabic-text-diacritization | Benchmark dataset with systems comparison |  |
🗣️ Dialect Identification
Resources for identifying and classifying Arabic dialects.
Shared Tasks
| Task | Description | Link |
|---|
| NADI 2025 | Multidialectal Arabic Speech Processing (8-way dialect + ASR) |  |
| NADI 2024 | Fifth Nuanced Arabic Dialect Identification |  |
| NADI Shared Tasks | Ongoing series of Arabic DID shared tasks |  |
Datasets
| Dataset | Description | Link |
|---|
| QADI | Twitter-based multi-class dialect classification |  |
| Arabic POS Dialect | POS tagging in Arabic dialects |  |
| Arabic Dialects to MSA | Parallel dialect-MSA corpus |  |
| Casablanca | Multidialectal Arabic speech dataset (NADI 2025) |  |
📊 Key Datasets
📝 Text Datasets
| Dataset | Description | Link |
|---|
| masader | Largest public catalogue of Arabic NLP datasets (600+) |  |
| 101 Billion Arabic Words | Massive Arabic web corpus |  |
| ArabicWeb24 | 39B+ tokens of high-quality Arabic web content |  |
| ArabicText-Large | 743K articles for LLM training |  |
| Arabic Billion Words | Abu El-Khair corpus: 5M+ articles, 1.5B+ words |  |
| Arabic Tweets | 41GB+ of Arabic tweets (~4B words) |  |
| Wojood | Nested NER corpus (550K tokens) |  |
| CIDAR | Culturally relevant instruction dataset (10K pairs) |  |
| Arabic_Function_Calling | First Arabic function calling dataset (50K+ samples) |  |
| ArabicaQA | Large-scale Arabic Question Answering |  |
| Mixed Arabic Datasets (MAD) | Community-driven diverse Arabic texts |  |
| Arabic-OpenHermes-2.5 | Arabic OpenHermes instruction dataset |  |
| Alpaca Arabic Instruct | Arabic Alpaca instruction dataset |  |
| Rasaif | Classical Arabic-English parallel texts (24 books) |  |
| Shifaa Medical | Arabic medical consultation dataset |  |
| Shifaa Mental Health | Arabic mental health consultations |  |
| Arabic Reasoning Dataset | 9.2K instruction-based reasoning QA pairs |  |
| Gazelle | Arabic writing assistance dataset |  |
| arabic-hate-speech-superset | Comprehensive hate speech detection |  |
| ArabicCorpus2B | 1.9B word Arabic corpus |  |
| The Arabic E-Book Corpus | 1,745 books (81.5M words) |  |
| BAREC Corpus | Arabic Readability Assessment |  |
| palm | Human-created Arabic instruction dataset |  |
| dialogue-arabic-dialects | Levantine, Egyptian, Gulf dialect dialogues |  |
| STMC | Saudi Tweets Mega Corpus (141M+ Saudi dialect tweets) |  |
| KFUPM Arabic AI Text Detection | Machine-generated Arabic text across multiple LLMs |  |
| Egyptian Dialogue | 4,322 Egyptian Arabic-English parallel pairs from TV subtitles |  |
| NileChat LHV-Egypt | Egyptian dialect dataset for LLM pre-training |  |
| NileChat Arabizi-Egypt | Egyptian Arabizi dataset for LLM pre-training |  |
🎤 Speech Datasets
| Dataset | Description | Link |
|---|
| ClArTTS | Classical Arabic TTS dataset |  |
| Arabic Speech Corpus | South Levantine Arabic (Damascian accent) |  |
| Arabic-English Code-Switching | Code-switching speech from YouTube |  |
| Egyptian Arabic ASR Clean | ~72 hours of Egyptian Arabic speech |  |
| MADIS5 | Spoken Arabic dialects |  |
| SADA | Saudi Audio Dataset: 668h from 57 TV shows, multi-dialect (SDAIA) |  |
| SADA22 (MSA) | MSA subset of SADA, Khaliji speech |  |
| SawtArabi | Arabic speech dataset |  |
🖼️ Vision & Multimodal Datasets
| Dataset | Description | Link |
|---|
| PEARL | Multimodal Culturally-Aware Arabic Instruction Dataset |  |
| Arabic-Image-Captioning_100M | 100 million Arabic image captions |  |
| Calliar | Online Arabic calligraphy (2500 samples) |  |
| arabic-img2md | 15K PDF pages paired with Markdown |  |
| Arabic-OCR-Dataset | 1M+ Arabic OCR samples |  |
| Arabic-VLM-Full-Pearl | 309K multimodal examples for VLM training |  |
| Tool | Description | Link |
|---|
| camel_tools | Suite of Arabic NLP tools (morphology, POS, NER, etc.) |  |
| Farasa | Fast and accurate Arabic text processing toolkit |  |
| SinaTools | Open source toolkit by SinaLab (Python APIs, CLI) |  |
| Qalsadi | Arabic morphological analyzer and lemmatizer |  |
| PyArabic | Python package for Arabic text manipulation |  |
| tnkeeh | Arabic text cleaning, normalization, preprocessing |  |
| Maha | Text processing library for Arabic text |  |
| Mishkal | Arabic text diacritizer (rule-based) |  |
| arabicprocess | Python library for Arabic preprocessing |  |
| MADAMIRA | Morphological analysis, diacritization, POS tagging |  |
📚 Specialized Libraries
| Library | Task | Link |
|---|
| klaam | Speech Recognition, Classification, TTS |  |
| tkseem | Arabic Tokenization |  |
| arabic-stop-words | Largest list of Arabic stop words |  |
| qawafi | Arabic poetry analysis |  |
| arabic_vocalizer | Deep-learning diacritization (ONNX format) |  |
🌍 Translation
| Tool | Description | Link |
|---|
| opus-mt-en-ar | English → Arabic neural MT |  |
| opus-mt-ar-en | Arabic → English neural MT (12.4M+ downloads) |  |
📄 Research Papers & Conferences
🎤 Conferences & Workshops
| Conference | Year | Link |
|---|
| ArabicNLP 2025 | 2025 |  |
| ArabicNLP 2024 | 2024 |  |
| ArabicNLP 2023 | 2023 |  |
| Arabic NLP Winter School | 2025 |  |
| AbjadNLP Workshop | 2025-2026 |  |
| OSACT | Ongoing |  |
📖 Foundational & Survey Papers
- The Landscape of Arabic Large Language Models (2025) [1]
- AraBERT: Transformer-based Model for Arabic Language Understanding (Antoun et al., 2020) [2]
- The Interplay of Variant, Size, and Task Type in Arabic Pre-trained Language Models (Inoue et al., 2021) [3]
- ARBERT & MARBERT: Deep Bidirectional Transformers for Arabic (Abdul-Mageed et al., 2021) [4]
- Jais and Jais-chat: Arabic-Centric Foundation and Instruction-Tuned LLMs (Sengupta et al., 2023) [5]
- Wojood: Nested Arabic Named Entity Corpus and Recognition (Jarrar et al., 2022) [6]
- Deep Learning for Arabic NLP: A Survey (Al-Ayyoub et al., 2018) [7]
- Evaluating Arabic LLMs: A Survey of Benchmarks, Methods, and Gaps (Alzubaidi et al., 2025) [8]
🆕 Recent Papers (2024-2025)
- Swan and ArabicMTEB — Dialect-Aware, Cross-Lingual Language Understanding (Bhatia et al., 2024) [9]
- GATE — General Arabic Text Embedding for Enhanced STS (Nacar et al., 2025) [10]
- A Survey of LLMs for Arabic Language and its Dialects (Mashaabi et al., 2024) [11]
- Hate speech detection in Arabic — corpus design and evaluation (2024) [12]
- NADI 2024 — Fifth Nuanced Arabic Dialect Identification Shared Task (2024) [13]
- CATT — Character-based Arabic Tashkeel Transformer (2024) [14]
- ArabianGPT — Native Arabic GPT-based LLM (2024) [15]
- SambaLingo — Teaching LLMs New Languages (2024) [16]
- SaudiBERT — BERT Pretrained on Saudi Dialect Corpora (2024) [17]
- ALLaM — Large Language Models for Arabic and English (SDAIA & IBM, 2024) [18]
- EgyBERT — BERT Pretrained on Egyptian Dialect Corpora (2024) [19]
🎓 Tutorials & Learning Resources
🏫 Academic Programs
| Resource | Description | Link |
|---|
| Arabic NLP Winter School | Two-day intensive at MBZUAI (Jan 2025) |  |
| ArabicNLP Conference | Annual ACL-affiliated conference by SIGARAB |  |
| AbjadNLP Workshop | NLP for languages using Arabic script |  |
| SDAIA Academy | Saudi AI & data science training programs |  |
| KAUST Academy (AI Specialization) | Free AI training for Saudi nationals |  |
💻 Online Resources
| Resource | Description | Link |
|---|
| Hugging Face NLP Course | Free NLP course (applicable to Arabic models) |  |
| AUB MIND Lab Arabic-NLP Demo | Interactive demo for Arabic NLP tasks |  |
| MoroccoAI Darija Resources | Curated Moroccan Arabic dialect NLP resources |  |
| NNLP-IL Arabic Resources | Comprehensive Arabic NLP resource list |  |
| 3ammiya | Egyptian Arabic dialect dictionary with 6,000+ entries |  |
🏛️ Research Institutions
Academic and government research labs driving Arabic NLP innovation.
| Institution | Country | Focus | Link |
|---|
| KAUST (CEMSE) | 🇸🇦 Saudi Arabia | Generative AI center, Arabic NLP research, sentiment analysis |  |
| SDAIA-KFUPM Joint Research Center (JRCAI) | 🇸🇦 Saudi Arabia | Arabic AI text detection, Arabic NLP datasets |  |
| Prince Sultan University (RIOTU Lab) | 🇸🇦 Saudi Arabia | ArabianGPT, Arabic IoT/robotics AI |  |
| King Saud University | 🇸🇦 Saudi Arabia | SaudiBERT, Saudi dialect corpora (STMC, SFC) |  |
| Nile University | 🇪🇬 Egypt | AI research, M.Sc. in AI co-designed with MIT/IBM |  |
| Ain Shams University | 🇪🇬 Egypt | Arabic NLP, sentiment analysis, NER research |  |
| MBZUAI | 🇦🇪 UAE | AIN, Atlas-Chat, ArabicMMLU, KITAB-Bench |  |
| CAMeL Lab (NYU Abu Dhabi) | 🇦🇪 UAE | CAMeLBERT, camel_tools, morphological analysis |  |
| UBC-NLP | 🇨🇦 Canada | MARBERT, AraT5, JASMINE, Swan, Palm dataset |  |
| AUB MIND Lab | 🇱🇧 Lebanon | AraBERT, AraGPT2, AraELECTRA |  |
| QCRI | 🇶🇦 Qatar | Farasa, Fanar LLM, Arabic POS tagging |  |
| SinaLab | 🇵🇸 Palestine | SinaTools, Wojood NER corpus |  |
| ARBML | 🌍 Community | masader (600+ datasets), klaam, tkseem, qawafi |  |
🏭 Companies & Startups
Companies and organizations building Arabic AI/NLP products and services.
🇦🇪 United Arab Emirates
| Company | Focus | Notable Products | Link |
|---|
| G42 | AI holding company, Arabic LLMs | Jais LLM, enterprise AI solutions |  |
| Inception AI | Arabic-centric foundation models | Jais model family (with Cerebras) |  |
| Technology Innovation Institute (TII) | Open-source LLMs, research | Falcon LLM family |  |
| Saal.ai | Cognitive AI solutions | Arabic NLP, speech, generative AI |  |
| Arabot | Conversational AI for Arabic | Arabic NLP chatbot engine |  |
| Arabic.AI (Tarjama) | Arabic-first autonomous AI | Pronoia Arabic LLM, Agentic AI platform |  |
🇸🇦 Saudi Arabia
| Company | Focus | Notable Products | Link |
|---|
| SDAIA | Sovereign AI, national data authority | ALLaM model, SADA dataset, NCAI, BALSAM benchmark |  |
| HUMAIN | PIF-backed full-stack AI company | ALLaM 34B, HUMAIN Chat, 8PB Arabic training data |  |
| Elm | Digital transformation, gov AI (PIF-backed) | Nuha Arabic LLM, legal AI assistant, gov platform |  |
| Misraj AI | Arabic-first AI ecosystem | Kawn LLM, Baseer OCR, Mutarjim, Workforces, SeamlessAPI |  |
| Mozn | Enterprise AI, Arabic NLU | OSOS Arabic NLU platform, FOCAL compliance suite |  |
| Unifonic | Conversational AI platform | Arabic-first CX Intelligence, AI chatbots |  |
| Lucidya | AI customer experience analytics | Arabic social listening, sentiment analysis |  |
| Hudhud AI | Arabic conversational AI (no-code SaaS) | Saudi-accent chatbots, Arabic-first customer engagement |  |
| Hazen.ai | AI traffic safety & computer vision | Deep learning road safety, seatbelt/phone detection |  |
| Future Look ITC (FLITC) | Arabic-native AI solutions, venture studio | LABEAH, Smart Hire, Rayee Media, Nabadat |  |
| NAMAA-Space | Arabic NLP models hub | Qari-OCR, EgypTalk-ASR, Masrawy translator, GLiNER Arabic NER |  |
| Wittify.ai | Conversational AI for Arabic | Interactive Arabic AI agents |  |
💡 Saudi Arabia's AI ecosystem: Saudi Arabia leads Arabic LLM development with 53+ Arabic language models identified by Q1 2025. Key initiatives include HUMAIN ($5B+ AWS partnership, NVIDIA AI factories), Vision 2030 AI Strategy (targeting 20,000 AI specialists, 300+ AI startups, $20B investment), and SDAIA Academy/SAMAI (training 1M citizens in AI).
🇪🇬 Egypt
| Company | Focus | Notable Products | Link |
|---|
| ITIDA / MCIT | National AI authority, sovereign models | Karnak LLM, SIA AI tutor, AcQua NLP, BelMasry, Torgoman |  |
| Intella | Arabic speech AI intelligence | Arabic STT across 25+ dialects (95.7% accuracy), Ziila digital human |  |
| WideBot AI | Arabic-first conversational AI | AQL Arabic LLM, chatbots, voicebots, AI agents |  |
| DXwand | Generative AI for Arabic business | ORXTRA platform, Arabic dialect chatbots, 20+ LLM support |  |
| Crowd Analyzer | Arabic social media monitoring | Arabic NLP analytics, sentiment analysis, media monitoring |  |
| Synapse Analytics | AI for financial inclusion | ML-powered credit scoring, Arabic data analytics |  |
| Convertedin | AI marketing automation | Arabic/English e-commerce personalization, $3M funded |  |
| Kngine | Semantic search & NLP | Arabic semantic search, data mining, knowledge engine |  |
| Monta AI | Enterprise AI solutions | LLM & RAG-based Arabic business automation |  |
| Elves | Conversational commerce | Arabic AI-assisted concierge, human-in-the-loop ML |  |
💡 Egypt's AI ecosystem: Egypt unveiled Karnak, its national LLM (30-80B params, highest-ranking Arabic LLM in its class), and launched the National AI Strategy 2025-2030 targeting $42.7B in AI economic value, 30,000 AI specialists, and 250+ AI startups. Karnak powers SIA (AI education tutor), AcQua (NLP call auditing), BelMasry (colloquial Arabic NLP engines), and Torgoman (specialized translation).
🌍 International (with Arabic focus)
| Company | Country | Focus | Notable Products | Link |
|---|
| SILMA AI | - | Arabic-first LLMs | SILMA LLMs, Arabic Broad Benchmark |  |
| Cohere | Canada | Multilingual LLMs | Command R Arabic, RAG optimization |  |
| Mistral AI | France | Multilingual LLMs | Mistral Saba (Arabic-optimized) |  |
| SambaNova Systems | USA | Arabic language adaptation | SambaLingo Arabic models |  |
| Mawdoo3 | Jordan / Saudi | Arabic AI & content, NLP toolkit | Arabic LLMs, largest Arabic website, Saudi expansion |  |
| Maqsam | Jordan / MENA | Arabic speech AI, call center AI | Arabic dialect STT, AI voice bots, surpasses Google/Microsoft |  |
📚 Awesome Lists
🤝 Contributing
Contributions are welcome! Please open an issue or submit a pull request to add new resources or update existing ones.
- Fork the repository
- Add your resource in the appropriate section
- Ensure links are valid and descriptions are concise
- Submit a pull request
📜 License
This work is licensed under a Creative Commons Attribution 4.0 International License.

📑 References
[1] The Landscape of Arabic Large Language Models. (2025). arXiv preprint arXiv:2506.01340. https://arxiv.org/html/2506.01340v1
[2] Antoun, W., Baly, F., & Hajj, H. (2020). AraBERT: Transformer-based Model for Arabic Language Understanding. OSACT. https://aclanthology.org/2020.osact-1.2/
[3] Inoue, G., et al. (2021). The Interplay of Variant, Size, and Task Type in Arabic Pre-trained Language Models. WANLP. https://aclanthology.org/2021.wanlp-1.10/
[4] Abdul-Mageed, M., et al. (2021). ARBERT & MARBERT: Deep Bidirectional Transformers for Arabic. ACL. https://aclanthology.org/2021.acl-long.551/
[5] Sengupta, N., et al. (2023). Jais and Jais-chat. arXiv:2308.16149. https://arxiv.org/abs/2308.16149
[6] Jarrar, M., et al. (2022). Wojood: Nested Arabic Named Entity Corpus. arXiv:2205.09651. https://arxiv.org/abs/2205.09651
[7] Al-Ayyoub, M., et al. (2018). Deep learning for Arabic NLP: A survey. Journal of Computational Science, 26. https://www.sciencedirect.com/science/article/pii/S1877750317303757
[8] Alzubaidi, A., et al. (2025). Evaluating Arabic LLMs: Benchmarks, Methods, and Gaps. arXiv:2510.13430. https://arxiv.org/abs/2510.13430
[9] Bhatia, G., et al. (2024). Swan and ArabicMTEB. arXiv:2411.01192. https://arxiv.org/abs/2411.01192
[10] Nacar, O., et al. (2025). GATE: General Arabic Text Embedding. arXiv:2505.24581. https://arxiv.org/abs/2505.24581
[11] Mashaabi, M., et al. (2024). A Survey of LLMs for Arabic. arXiv:2410.20238. https://arxiv.org/abs/2410.20238
[12] Hate speech detection in Arabic. (2024). Frontiers in AI. https://www.frontiersin.org/journals/artificial-intelligence/articles/10.3389/frai.2024.1345445/full
[13] NADI 2024: Fifth Nuanced Arabic Dialect Identification. (2024). ArabicNLP. https://arxiv.org/abs/2407.04910
[14] CATT: Character-based Arabic Tashkeel Transformer. (2024). arXiv:2407.03236. https://arxiv.org/abs/2407.03236
[15] ArabianGPT: Native Arabic GPT-based LLM. (2024). arXiv:2402.15313. https://arxiv.org/abs/2402.15313
[16] SambaLingo: Teaching LLMs New Languages. (2024). arXiv:2404.05829. https://arxiv.org/abs/2404.05829
[17] SaudiBERT: A Large Language Model Pretrained on Saudi Dialect Corpora. (2024). arXiv:2405.06239. https://arxiv.org/abs/2405.06239
[18] ALLaM: Large Language Models for Arabic and English. (2024). arXiv:2407.15390. https://arxiv.org/abs/2407.15390
[19] EgyBERT: A Large Language Model Pretrained on Egyptian Dialect Corpora. (2024). arXiv:2408.03524. https://arxiv.org/abs/2408.03524
[20] Multilingual E5 Text Embeddings: A Technical Report. (2024). arXiv:2402.05672. https://arxiv.org/abs/2402.05672
If you find this resource helpful, please give it a ⭐
Made with ❤️ for the Arabic NLP community