McGill-NLP/AfriqueQwen3.5-9B-50Langs-Instruct-v1

VISIONPricing:Input $0.431 / Cached $0.0862 / Output $1.12Concurrent Unit Cost:1Model Size:9BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Sep 11, 2026License:cc-by-4.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

McGill-NLP/AfriqueQwen3.5-9B-50Langs-Instruct-v1 is a 9 billion parameter instruction-tuned language model developed by McGill-NLP, based on the Qwen3.5 architecture. This model is specifically optimized for performance across 50 African languages, demonstrating strong capabilities in tasks like mathematics, common sense reasoning, and translation. It features a 32768 token context length and achieves an overall score of 68.3 on a suite of African language benchmarks, making it suitable for multilingual applications in African contexts.

Loading preview...

AfriqueQwen3.5-9B-50Langs-Instruct-v1 Overview

This model is an instruction-tuned version of the McGill-NLP/AfriqueQwen3.5-9B-50Langs base model, specifically designed to excel in a wide array of tasks across 50 African languages. With 9 billion parameters and a substantial context length of 32768 tokens, it offers robust performance for complex multilingual applications.

Key Capabilities

  • Multilingual Proficiency: Demonstrates strong performance across various African language benchmarks, including AfriMGSM (mathematics), AfriMMLU (common sense reasoning), AfriXNLI (natural language inference), Belebele, Injongo, SIB-200, and FLORES (translation).
  • Instruction Following: Fine-tuned to follow instructions effectively, as evidenced by its Instruct-v1 designation.
  • Mathematical Reasoning: Achieves a score of 64.6 on AfriMGSM, indicating strong capabilities in solving math word problems, with specific prompting recommendations for step-by-step reasoning.
  • High Benchmark Scores: Outperforms several other models in its class, including Gemma-3-4B-it and Qwen3.5-4B, with an overall score of 68.3 on the comprehensive African language benchmark suite.

Good For

  • African Language Applications: Ideal for developers building applications that require high-quality language understanding and generation in diverse African linguistic contexts.
  • Multilingual Instruction Following: Suitable for tasks requiring precise instruction adherence across multiple languages.
  • Mathematical Problem Solving: Can be effectively used for solving mathematical word problems, especially when guided by the recommended prompting strategies.
  • Translation and NLI: Strong performance on AfriXNLI and FLORES benchmarks suggests its utility in translation and natural language inference tasks for African languages.

Finetunes
1 models