LenguajeNatural.AI LeNIA-Chat Qwen2-1.5B-v0

Developed by LenguajeNatural.AI, this 1.5 billion parameter model is the second iteration in their LeNIA series, built upon the Qwen2 architecture. It is specifically designed and extensively trained for the Spanish-speaking community, focusing on advanced text generation, chat, and instruction-following tasks.

Key Capabilities & Features

• Spanish-Exclusive Training: The model has been trained exclusively in Spanish to maximize effectiveness for Spanish-speaking users.
• Optimized for Chat & Instructions: Training involved three distinct phases: domain-specific FLAN-style adaptation, high-quality instruction tuning, and chat/abstractive QA fine-tuning.
• Extended Context Window: Features a maximum sequence length of 32,768 tokens, a significant increase from previous versions.
• Improved Performance: Despite being smaller (1.5B parameters) than its predecessor, it significantly outperforms it.

Use Cases & Limitations

This model is ideal for applications requiring text generation, chatbots, and virtual assistants in Spanish. Users should be aware of the inherent limitations of language models, especially those of its size (1.5B parameters), and evaluate its performance within their specific use context. It is distributed under the Apache 2.0 License.