The PazaBench Leaderboard is an Automatic Speech Recognition (ASR) benchmark for low-resource languages developed by the Microsoft Research Africa, Nairobi Lab. In the latest release, PazaBench covers 61 African Languages across 53 State-of-the-Art ASR and Language Models. PazaBench compares three key metrics: Character Error Rate (CER), Word Error Rate (WER), and RTFx (Inverse Real-Time Factor).
Compare models at a glance. Use the filters below to customize your view or explore the data directly in the tables.
💡 Leaderboard Guide: Languages are ordered alphabetically. Model families are ranked from left to right by the average performance across languages. Click on filter & customize above to filter languages or model.
📉 Lower is better — Character Error Rate measures the percentage of characters incorrectly transcribed. This is especially important for languages with rich word forms, where meaning is built by combining word parts, therefore errors at the character level can significantly impact meaning.
See 📊 Visualizations tab for CER Performance chart.