Haiduk-27B: A Multilingual Instruction-Tuned Model

Haiduk-27B is a 27 billion parameter instruction-tuned language model developed by iuriesula99, building upon Google's Gemma-3-27b-it and TheDrummer/Big-Tiger-Gemma-27B-v3. It features a substantial context window of 32768 tokens, enabling it to process and generate longer, more coherent texts.

Key Capabilities

• Multilingual Support: Trained with a focus on English, Romanian, and Russian, making it suitable for applications requiring proficiency in these languages.
• Extended Context Window: The 32768-token context length allows for handling complex queries, detailed conversations, and extensive document analysis.
• Instruction Following: As an instruction-tuned model, Haiduk-27B is designed to accurately interpret and execute user commands, making it versatile for various NLP tasks.

Training Data

The model's training incorporates several specialized datasets:

• iuriesula99/TigerPhase1
• iuriesula99/TigerDataset2
• iuriesula99/TigerPhase3
• iuriesula99/TigerPhase4

Good For

• Applications requiring strong performance in English, Romanian, and Russian.
• Tasks benefiting from a large context window, such as summarization of long documents, detailed question answering, and complex code generation.
• General instruction-following tasks where precise and context-aware responses are crucial.