smshahbaj/Rifa-Nano-0.5B

TEXT GENERATIONPricing:Input $0.04 / Cached $0.008 / Output $0.08Concurrent Unit Cost:1Model Size:0.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Aug 26, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

Rifa-Nano-0.5B is a 0.5 billion parameter instruction-tuned language model developed by SM Shahbaj, based on Qwen2.5-0.5B-Instruct. It is designed as a lightweight, helpful assistant with strong support for the Bangla language. A key differentiator is its dedicated anti-hallucination training, aimed at reducing confident wrong answers on unanswerable questions. This model is optimized for general knowledge, light math, coding, and particularly for applications requiring efficient Bangla language processing.

Loading preview...

RIFA Nano (0.5B) Overview

RIFA Nano is the smallest model in the RIFA series, developed and fine-tuned by SM Shahbaj. It is based on Qwen/Qwen2.5-0.5B-Instruct and features 0.5 billion parameters, making it a lightweight option for various applications.

Key Capabilities and Features

  • Multilingual Support: Strong emphasis on Bangla language capabilities, alongside English.
  • Reduced Hallucination: Incorporates dedicated anti-hallucination training to minimize confident incorrect responses to unanswerable questions.
  • Balanced Skill Mix: Offers a blend of general knowledge, light mathematical reasoning, and coding abilities.
  • Lightweight Design: As a 0.5B parameter model, it is suitable for resource-constrained environments or applications requiring fast inference.
  • GGUF Quantizations: Available in various GGUF formats (e.g., Q3_K_M, Q4_K_M, Q5_K_M, Q6_K, Q8_0) for efficient deployment on consumer hardware, with Q5_K_M recommended for quality and size balance.

Ideal Use Cases

RIFA Nano is well-suited for:

  • Applications requiring a compact, efficient language model.
  • Tasks involving the Bangla language, where its specialized training provides an advantage.
  • Use cases where reducing model hallucination on unknown queries is critical.
  • As a foundational assistant for general queries, light coding tasks, and basic math problems, especially when deployed on edge devices or with limited computational resources.