Dicta-il/DictaLM-3.0-1.7B-Base is a 1.7 billion parameter base language model developed by Dicta, initialized from Qwen3-1.7B-Base. Trained on extensive Hebrew and English corpora, it establishes a new state-of-the-art for Hebrew LLMs in its weight class. This model is designed for further fine-tuning and excels in tasks requiring strong Hebrew language understanding and generation.
Loading preview...
Dicta-LM 3.0: Hebrew Sovereign LLM
Dicta-LM 3.0 is an open-weight collection of large language models developed by Dicta, specifically designed to advance the frontier of Hebrew LLMs. This particular model, DictaLM-3.0-1.7B-Base, is a 1.7 billion parameter base model initialized from Qwen3-1.7B-Base, available in full precision (BF16).
Key Capabilities & Features
- State-of-the-Art Hebrew Performance: Sets a new benchmark for its weight class in Hebrew language tasks, both as a base model and for subsequent chat model fine-tuning.
- Extensive Training Data: Trained on large corpora of both Hebrew and English texts, ensuring strong bilingual capabilities with a focus on Hebrew.
- Base Model Design: Provided as a foundational model, ideal for developers to fine-tune for specific downstream applications and use cases.
- Open-Weight & Unlimited Use: Available for download and unlimited use, promoting accessibility and further research in Hebrew NLP.
Intended Use Cases
- Fine-tuning: Excellent starting point for developing custom Hebrew-centric applications, such as chatbots, content generation, or language understanding systems.
- Research & Development: Suitable for researchers exploring new methods in Hebrew natural language processing and model adaptation.
Important Note
This is a base model and does not include built-in moderation mechanisms. It is not an instruction-tuned chat model; however, chat variants are available within the DictaLM 3.0 collection.