swapnillo/Bangla-AI-1.7B

Hugging Face
TEXT GENERATIONPricing:Input $0.32 / Cached $0.064 / Output $1.6Concurrent Unit Cost:1Model Size:2BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Nov 5, 2025License:mitArchitecture:Transformer Open Weights Featherless Exclusive Warm

swapnillo/Bangla-AI-1.7B is a 1.7 billion parameter causal language model, fine-tuned from Qwen3-1.7B by Ismam Nur Swapnil. Optimized for Bengali (Bangla) language, it excels at understanding and generating natural, culturally appropriate Bengali responses. This model is specifically designed for conversational AI applications, chatbots, and instruction-following tasks in Bengali, leveraging LoRA fine-tuning on a 100K Bengali instruction dataset.

Loading preview...

Overview

swapnillo/Bangla-AI-1.7B is a specialized large language model developed by Ismam Nur Swapnil, fine-tuned from the Qwen3-1.7B base model. It focuses exclusively on the Bengali (Bangla) language, aiming to provide accurate, culturally appropriate, and grammatically correct responses. The model utilizes LoRA (Low-Rank Adaptation) for efficient fine-tuning on a dedicated 100,000-pair Bengali instruction dataset, making it highly proficient in Bengali conversational AI.

Key Capabilities

  • Bengali Language Generation: Optimized for natural and fluent Bengali text generation.
  • Instruction Following: Capable of understanding and executing instructions given in Bengali.
  • Cultural Appropriateness: Designed to produce responses that align with Bengali cultural norms.
  • Grammatical Accuracy: Focuses on proper Bengali grammar and conversational style.
  • Efficient Fine-tuning: Built with LoRA on a 1.7 billion parameter base, allowing for specialized performance without the overhead of larger models.

Use Cases

This model is ideal for applications requiring strong Bengali language capabilities:

  • Bengali Chatbots and Virtual Assistants: Creating interactive agents for Bengali-speaking users.
  • Question-Answering Systems: Developing systems that can answer queries in Bengali.
  • Content Generation: Producing various forms of text content in Bengali.
  • Further Fine-tuning: Can serve as a base for domain-specific Bengali NLP tasks, such as customer service or educational tools.

Limitations

Users should be aware that the model's performance is tied to its training data, potentially leading to factual inaccuracies or biases. It is not recommended for high-stakes decision-making or for use in languages other than Bengali.

Finetunes
1 models