adzcai/AfriGuardInst-AfriqueQwen3.5-4B-50Langs-Instruct-v1

VISIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:4.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Sep 28, 2026License:cc-by-4.0Architecture:Transformer Open Weights Featherless Exclusive Cold

AfriGuardInst-AfriqueQwen3.5-4B-50Langs-Instruct-v1 is a 4.5 billion parameter instruction-tuned causal language model developed by adzcai. It is a fine-tuned version of McGill-NLP/AfriqueQwen3.5-4B-50Langs-Instruct-v1, specifically optimized for safety system instructions and response generation using the AfriGuard-inst dataset. This model excels at generating safety-aligned responses, making it suitable for applications requiring content moderation and responsible AI outputs.

Loading preview...

Overview

AfriGuardInst-AfriqueQwen3.5-4B-50Langs-Instruct-v1 is a 4.5 billion parameter instruction-tuned model developed by adzcai. It is built upon the McGill-NLP/AfriqueQwen3.5-4B-50Langs-Instruct-v1 base model and has been fine-tuned using the israel/AfriGuard-inst dataset. The fine-tuning process focused on generating safety-aligned responses, where each example includes safety system instructions, a user prompt, and a target output structured with <safety>, <category>, and <response> tags.

Key Capabilities

  • Safety-aligned response generation: Specifically trained to produce responses that adhere to defined safety guidelines and categories.
  • Instruction following: Designed to interpret and act upon safety system instructions provided in the prompt.
  • Multilingual foundation: Benefits from the multilingual capabilities of its AfriqueQwen3.5-4B-50Langs-Instruct-v1 base model.

Training Details

The model was fine-tuned for 1 epoch using DeepSpeed ZeRO-3 with a learning rate of 1e-05 and a total batch size of 2. The training objective focused on minimizing loss on the generated response portion of the output. This model serves as a matched baseline for adzcai/AfriGuardPlain-AfriqueQwen3.5-4B-50Langs-Instruct-v1.

Good For

  • Content moderation systems: Automating the detection and generation of safe responses.
  • Responsible AI applications: Building systems that require explicit safety alignment in their outputs.
  • Research in AI safety: Investigating instruction-tuned models for safety and moderation tasks.

Finetunes
2 models