Hebrew AI Models
April 6, 2026 · View on GitHub
Hebrew AI Models
A catalog of AI models for the Hebrew language — LLMs, speech-to-text, text-to-speech, NLP, embeddings, translation, and more.
This is a community resource. Contributions welcome via pull request.
Last updated: April 1, 2026
Table of Contents
- Large Language Models (LLMs)
- Speech-to-Text / ASR
- Text-to-Speech (TTS)
- NLP Models
- Embeddings
- Translation
- Summarization
- OCR / Vision
- Speech Foundation Models
- Curated HF Collections
- Notable Projects & Resources
- Key Organizations
Large Language Models (LLMs)
dicta-il (Israel Center for Text Analysis)
| Model | Size | Type | Description |
|---|---|---|---|
| DictaLM-3.0-24B-Base | 24B | Base | Latest generation Hebrew LLM, large variant |
| DictaLM-3.0-24B-Thinking | 24B | Reasoning | Thinking/reasoning variant |
| DictaLM-3.0-Nemotron-12B-Base | 12B | Base | Nemotron-based Hebrew LLM |
| DictaLM-3.0-Nemotron-12B-Instruct | 12B | Instruct | Instruction-tuned Nemotron |
| DictaLM-3.0-1.7B-Base | 1.7B | Base | Compact Hebrew LLM |
| DictaLM-3.0-1.7B-Instruct | 1.7B | Instruct | Instruction-tuned compact model |
| DictaLM-3.0-1.7B-Thinking | 1.7B | Reasoning | Thinking variant of compact model |
| dictalm2.0 | 7B | Base | Previous-gen Hebrew LLM (Mistral-based) |
| dictalm2.0-instruct | 7B | Instruct | Instruction-tuned DictaLM 2.0 |
| BEREL_3.0 | — | Base | Biblical/Rabbinic Hebrew language model |
| BEREL | — | Base | BERT for Early/Rabbinic Hebrew |
yam-peleg
| Model | Size | Type | Description |
|---|---|---|---|
| Hebrew-Gemma-11B-V2 | 11B | Chat | Hebrew-adapted Gemma, v2 |
| Hebrew-Gemma-11B-Instruct | 11B | Instruct | Instruction-tuned Hebrew Gemma |
| Hebrew-Gemma-11B | 11B | Base | Base Hebrew Gemma |
| Hebrew-Mistral-7B | 7B | Base | Hebrew-adapted Mistral 7B |
| Hebrew-Mistral-7B-200K | 7B | Base | Extended context (200K) Hebrew Mistral |
Other Hebrew LLMs
| Model | Author | Size | Type | Description |
|---|---|---|---|---|
| Hebrew_Nemo | SicariusSicariiStuff | 12B | Instruct | Hebrew fine-tune of Mistral Nemo |
| Hebrew-GPT | XythicK | 1B | Instruct | Hebrew fine-tune of Llama 3.2 1B |
| HebrewGPT-1B | Slasky | 1B | Base | Hebrew-native GPT trained from scratch |
| HebrewGPT-296M | Slasky | 296M | Base | Smaller Hebrew-native GPT |
| Hebrew-GPT2-345M-Stage | Norod78 | 345M | Base | Hebrew GPT-2 |
| hebrew-gpt_neo-small | Norod78 | ~125M | Base | Hebrew GPT-Neo small |
| Llama-3.2-3B-Hebrew-Master | yonioz123 | 3B | Chat | Hebrew fine-tune of Llama 3.2 |
| hebrew-math-tutor-v1 | Intel | 4B | Instruct | Hebrew math tutoring model |
| Hebrew-Mistral-7B-Instruct-v0.1 | ronmasas | 7B | Instruct | Hebrew instruction-tuned Mistral |
| dictalm2.0-instruct-fine-tuned-alpaca-gpt4-hebrew | ronigold | 7B | Instruct | DictaLM 2.0 fine-tuned on Alpaca GPT-4 Hebrew |
| SmolLM-135M-FakyPedia-EngHeb | Norod78 | 135M | Base | Bilingual English-Hebrew small LM |
| gemma-3_4b_hebrew-lyrics-finetune | Norod78 | 4B | Fine-tune | Gemma 3 fine-tuned for Hebrew lyrics |
Speech-to-Text / ASR
ivrit-ai
The leading Hebrew ASR project. Community-driven.
| Model | Size | Type | Description |
|---|---|---|---|
| whisper-large-v3-turbo-ct2 | Large | CTranslate2 | Top Hebrew ASR model, CT2 optimized (22K+ downloads) |
| whisper-large-v3 | Large-v3 | Fine-tuned | Hebrew fine-tuned Whisper large-v3 |
| whisper-large-v3-turbo | Large-v3 | Fine-tuned | Hebrew fine-tuned Whisper turbo |
| whisper-large-v3-ct2 | Large-v3 | CTranslate2 | CT2 format of Hebrew Whisper |
| whisper-v2-d4 | Large-v2 | Fine-tuned | Earlier Hebrew Whisper fine-tune |
| faster-whisper-v2-d4 | Large-v2 | Faster-Whisper | Faster-Whisper format |
| pyannote-speaker-diarization-3.1 | — | Diarization | Hebrew speaker diarization |
| pyannote-segmentation-3.0 | — | Segmentation | Hebrew audio segmentation |
GitHub: ivrit-ai/ivrit.ai | ivrit-ai/asr-training
imvladikon
| Model | Size | Type | Description |
|---|---|---|---|
| wav2vec2-xls-r-300m-hebrew | 300M | ASR | Most downloaded Hebrew model (144K downloads) |
| wav2vec2-large-xlsr-53-hebrew | 300M | ASR | Hebrew fine-tuned XLSR-53 |
| wav2vec2-xls-r-1b-hebrew | 1B | ASR | Hebrew fine-tuned XLS-R 1B |
| wav2vec2-xls-r-300m-lm-hebrew | 300M | ASR+LM | With language model decoding |
| whisper-medium-he | Medium | ASR | Hebrew Whisper medium |
Other ASR
| Model | Author | Description |
|---|---|---|
| whisper-heb-ipa | thewh1teagle | Hebrew Whisper with IPA phoneme output |
| WhiStress | slprl | Hebrew stress/prosody-aware ASR |
| hebrew_whisper | ShmuelRonen | Hebrew Whisper ASR project |
Text-to-Speech (TTS)
An emerging area for Hebrew — no production-grade standalone model yet, but active research.
| Model / Project | Author | Description |
|---|---|---|
| HebTTS | slp-rl (Technion) | Diacritic-free Hebrew TTS research |
| SIMS-Llama3.2-3B | slprl | Speech-language model, 3B |
| SIMS-7B | slprl | Larger speech-language model |
| israwave | thewh1teagle | Hebrew TTS project |
| Zonos-Hebrew | notmax123 | Hebrew variant of Zonos TTS |
| hebrewTTS-orpheus3b | Trojanssafsdg | Hebrew Orpheus TTS 3B |
| hebrew-orpheus-tts | rotem1121 | Hebrew Orpheus TTS variant |
| phonikud | thewh1teagle | Hebrew diacritization/phonemization for TTS pipelines |
| mms-tts-heb | thewh1teagle | Hebrew MMS TTS |
NLP Models
BERT-based Foundation Models
| Model | Author | Description |
|---|---|---|
| AlephBERT-base | onlplab (BIU) | Foundational Hebrew BERT (29.6K downloads) |
| heBERT | avichr | Hebrew BERT |
| Legal-heBERT | avichr | Hebrew BERT for legal domain |
| DictaBERT | dicta-il | Hebrew BERT from Dicta |
| DictaBERT-large | dicta-il | Large Hebrew BERT |
| NeoDictaBERT | dicta-il | Next-gen DictaBERT |
| NeoDictaBERT-bilingual | dicta-il | Bilingual Hebrew-English BERT |
| AlephBertGimmel-base | dicta-il | Third-generation Hebrew BERT |
| MsBERT | dicta-il | Hebrew BERT variant |
| HeRo | HeNLP | Hebrew RoBERTa |
| LongHeRo | HeNLP | Long-context Hebrew RoBERTa |
Named Entity Recognition (NER)
| Model | Author | Description |
|---|---|---|
| heBERT_NER | avichr | Hebrew NER (39K downloads) |
| dictabert-large-ner | dicta-il | Large Hebrew NER |
| dictabert-ner | dicta-il | Standard Hebrew NER |
Sentiment Analysis
| Model | Author | Description |
|---|---|---|
| heBERT_sentiment_analysis | avichr | Hebrew sentiment analysis |
| dictabert-sentiment | dicta-il | Hebrew sentiment classifier |
| hebEMO | avichr | Hebrew emotion detection (8 emotions: anger, fear, joy, surprise, sadness, disgust, anticipation, trust) |
Morphology & Parsing
| Model | Author | Description |
|---|---|---|
| dictabert-morph | dicta-il | Hebrew morphological analysis |
| dictabert-joint | dicta-il | Joint morphological analysis |
| dictabert-large-parse | dicta-il | Hebrew syntactic parsing |
| dictabert-parse | dicta-il | Standard Hebrew parser |
| dictabert-seg | dicta-il | Hebrew word segmentation |
| dictabert-lex | dicta-il | Hebrew lexical analysis |
| dictabert-syntax | dicta-il | Hebrew syntax analysis |
Diacritization (Nikud)
| Model | Author | Description |
|---|---|---|
| dictabert-large-char-menaked | dicta-il | Hebrew diacritization / nikud restoration (4.5K downloads) |
| dictabert-char | dicta-il | Character-level Hebrew model |
Other NLP
| Model | Author | Type | Description |
|---|---|---|---|
| hebrew_punctuation | Verbit | Punctuation | Hebrew punctuation restoration |
| dictabert-heq | dicta-il | QA | Hebrew extractive question answering |
| hebert_parashoot | imvladikon | QA | Hebrew reading comprehension |
| hebrew-offensive-detection | KevynKrancenblum | Classification | Hebrew offensive language detection |
| hebert-finetuned-hebrew-metaphor | tdklab | Classification | Hebrew metaphor detection |
Embeddings
| Model | Author | Description |
|---|---|---|
| neodictabert-bilingual-embed | dicta-il | Bilingual Hebrew-English embeddings |
| sentence-transformers-alephbert | imvladikon | Hebrew sentence embeddings (4.9K downloads) |
| sentence_transformers_alephbertgimmel_small | imvladikon | Smaller Hebrew sentence embeddings |
Translation
| Model | Author | Direction | Description |
|---|---|---|---|
| opus-mt-en-he | Helsinki-NLP | EN → HE | English to Hebrew translation |
| opus-mt-tc-big-he-en | Helsinki-NLP | HE → EN | Hebrew to English, large model |
| english-hebrew-translation | ashercn97 | EN ↔ HE | T5-based English-Hebrew translation |
Summarization
| Model | Author | Description |
|---|---|---|
| het5_summarization | imvladikon | Hebrew T5 for summarization |
| het5_small_summarization | imvladikon | Smaller Hebrew T5 summarization |
| cross_summarization_he_en | imvladikon | Cross-lingual Hebrew-English summarization |
OCR / Vision
| Model | Author | Description |
|---|---|---|
| testing-trOCR-hebrew-handwritten | sivan22 | TrOCR for Hebrew handwritten text |
| paleo-hebrew-mt5-post-ocr-processing | mr3vial | Post-OCR correction for paleo-Hebrew |
| paleo-hebrew-qwen3-vl-lora-post-ocr-processing | mr3vial | Vision-language post-OCR for paleo-Hebrew |
Speech Foundation Models
| Model | Author | Description |
|---|---|---|
| mhubert-base-25hz | slprl (Technion) | mHuBERT speech features at 25Hz |
| slam / slam_scaled / slam_large | slprl | Speech-language alignment models |
| StresSLM | slprl | Hebrew stress-aware speech-language model |
| PAST | slprl | Prosody-aware speech model |
Curated HF Collections
Curated Hugging Face collections (by danielrosehill) grouping Hebrew models, datasets, and spaces by task:
| Collection | Description |
|---|---|
| Hebrew Large Language Models | LLMs with Hebrew capabilities |
| Hebrew ASR | Hebrew automatic speech recognition models |
| Whisper Hebrish | Whisper-based Hebrew/English models |
| Hebrew TTS | Hebrew text-to-speech models |
| English-Hebrew Translation | English ↔ Hebrew translation models |
| Hebrew Punctuation Restoration | Models for restoring punctuation in Hebrew text |
| Hebrew Diacritic Restoration | Models for restoring nikud/diacritics |
| Hebrew Sentiment Classification Models | Sentiment classification |
| Hebrew OCR Models | Hebrew OCR and vision models |
| Hebrew Datasets | Datasets for Hebrew NLP/AI |
| Hebrew AI Spaces | Hugging Face Spaces for Hebrew AI |
Notable Projects & Resources
| Project | Description |
|---|---|
| NNLP-IL/Hebrew-Resources | Comprehensive list of Hebrew NLP resources |
| iddoberger/awesome-hebrew-nlp | Curated list of Hebrew NLP resources |
| ivrit-ai/ivrit.ai | Main ivrit.ai codebase for Hebrew ASR |
| ivrit-ai/asr-training | ASR training recipes for Hebrew |
| ivrit-ai/eliezer | ivrit.ai's Hebrew bot |
| slp-rl/HebTTS | Diacritic-free Hebrew TTS research |
| thewh1teagle/israwave | Hebrew TTS project |
| HebrewNLP/HebrewNLP4J | Java library for HebrewNLP API |
| danielrosehill/Hebrew-TTS-Providers | Directory of Hebrew TTS providers and services |
| danielrosehill/Hebrew-Image-Generation-Eval | Evaluation of image generation models for Hebrew text |
| danielrosehill/Hebrew-LLMs | Resource list of Hebrew large language models |
| AvrahamRaviv/Deep-Learning-in-Hebrew | Deep learning educational resources in Hebrew |
| ShmuelRonen/hebrew_whisper | Hebrew Whisper ASR project |
Key Organizations
| Organization | Focus | Notes |
|---|---|---|
| dicta-il | LLMs, BERT, NLP | Israel Center for Text Analysis. Most comprehensive Hebrew NLP toolkit |
| ivrit-ai | ASR / STT | Leading Hebrew speech recognition project. Community-driven. Website: ivrit.ai |
| yam-peleg | LLMs | Hebrew Gemma and Mistral adaptations |
| avichr | NLP | heBERT family — widely used Hebrew NER and sentiment |
| onlplab | Foundation BERT | AlephBERT — foundational Hebrew BERT from Bar-Ilan University |
| imvladikon | ASR, embeddings | wav2vec2 Hebrew (144K downloads), sentence transformers |
| slprl | TTS, speech | Technion Speech Processing Lab |
| HeNLP | Foundation models | HeRo (Hebrew RoBERTa) |
| Helsinki-NLP | Translation | OPUS-MT Hebrew translation models |
| Norod78 | Small LLMs | Hebrew GPT-2, GPT-Neo, creative fine-tunes |
| Slasky | LLMs | Hebrew-native GPTs trained from scratch |
| thewh1teagle | TTS, phonemization | israwave TTS, phonikud diacritization |
| Intel | Education LLMs | Hebrew math tutoring model |
Contributing
Know of a Hebrew AI model not listed here? Open a pull request or file an issue.
License
This work is licensed under a Creative Commons Attribution 4.0 International License.