McGill-NLP/AfriqueQwen-14B

Model Overview

McGill-NLP/AfriqueQwen-14B is a 14 billion parameter causal language model, part of the AfriqueLLM suite, developed by McGill-NLP. It is built upon the Qwen3-14B-Base architecture and has undergone extensive continued pre-training (CPT) on approximately 26 billion tokens of multilingual data, specifically targeting 20 African languages. This adaptation significantly enhances its performance in these languages while preserving capabilities in high-resource languages like English and French.

Key Capabilities

• Multilingual Proficiency: Adapted for 20 African languages including Swahili, Hausa, Yoruba, and Amharic, alongside strong performance in high-resource languages.
• Robust Base: Leverages the Qwen 3 14B Base model, known for strong performance and long-context handling (32,768 tokens native context length).
• Specialized Training: Continued pre-training on a diverse corpus including African monolingual data (22.8B tokens), code (1B tokens), and mathematics (~1B tokens) to ensure broad utility.
• Performance Gains: Achieves significant improvements on African language benchmarks, with a +23.3 (57.9%) overall score increase on the primary evaluation suite compared to its base model.

Good For

• Applications requiring high-quality language understanding and generation in African languages.
• Research and development focusing on multilingual NLP, especially for under-resourced languages.
• Tasks benefiting from a long context window in a multilingual setting.