DictaLM-3.0-1.7B-Thinking: A Hebrew Reasoning Chat Model
DictaLM-3.0-1.7B-Thinking is part of the Dicta-LM 3.0 collection, an open-weight suite of large language models developed by Dicta. This specific model is a 1.7 billion parameter reasoning chat model, initialized from Qwen3-1.7B-Base, and is notable for its full precision (BF16).
Key Capabilities & Features
- Reasoning Chat Model: Before generating a response, the model utilizes a designated 'thinking block' to plan its reply, enhancing the coherence and quality of its output.
- Hebrew Language Focus: It sets a new state-of-the-art (SOTA) for its weight class in Hebrew language processing, both as a base and chat model.
- Tool-Calling Support: The model supports tool-calling, enabling integration with external tools and APIs, as demonstrated with vLLM.
- Extensive Training: Trained on comprehensive corpora of both Hebrew and English texts.
When to Use This Model
- Hebrew-centric Applications: Ideal for chatbots, conversational AI, and other language generation tasks where strong Hebrew performance is critical.
- Reasoning-intensive Tasks: Suitable for use cases that benefit from a model's ability to internally plan and structure its responses.
- Integration with External Tools: Leverage its tool-calling capabilities for applications requiring interaction with APIs or other software components.