alan-turing-institute/t0-1.1-k5-1.5B

TEXT GENERATIONPricing:Input $0.04 / Cached $0.008 / Output $0.08Concurrent Unit Cost:1Model Size:1.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Jun 8, 2025License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

t0-1.1-k5-1.5B is a 1.5 billion parameter causal language model developed by the Alan Turing Institute, fine-tuned from Qwen2.5-1.5B-Instruct. It is specifically optimized for Retrieval-Augmented Generation (RAG) pipelines, particularly for question answering over domain-specific knowledge bases. This model excels as a generator component in RAG systems, demonstrated with health-related information from NHS A-to-Z condition webpages, utilizing top-k=5 retrieval.

Loading preview...

Model Overview

t0-1.1-k5-1.5B is a 1.5 billion parameter language model developed by the Alan Turing Institute, stemming from their t0 research initiative focused on lean yet capable LLMs for Retrieval-Augmented Reasoning (RAR). Fine-tuned from Qwen2.5-1.5B-Instruct, this model is specifically optimized for integration into Retrieval-Augmented Generation (RAG) pipelines.

Key Capabilities and Features

  • RAG Optimization: Designed to function as the generator component within a RAG system, reasoning over retrieved documents from a domain-specific knowledge base.
  • Domain-Specific Application: Demonstrated effectiveness in health-related question answering, using NHS A-to-Z condition webpages as a reference knowledge source.
  • Lean Architecture: As a 1.5B parameter model, it aims for efficiency while maintaining strong performance in its intended RAG context.
  • English Language Support: Primarily intended for English-language text generation and conversational tasks.

Intended Use Cases

This model is best suited for:

  • Text generation and conversational tasks within a RAG framework.
  • Question answering where information is retrieved from a specific corpus.
  • Integration into RAG systems for domain-specific knowledge bases, such as health information portals.

Limitations and Considerations

Users should be aware that the model:

  • Inherits biases from its base model and training data (e.g., NHS A-to-Z corpus).
  • Is not intended for medical diagnosis or clinical decision-making without professional oversight.
  • May be prone to errors or hallucinations due to its smaller parameter count compared to larger models.

For more details, refer to the t0-1 repository and the associated research paper.

Finetunes
1 models