ResearchBenchmarksTokenization Efficiency on Bengali (bn)Follow6.05Bytes per TokenLLaMA-45.8747.0628.259.438Nov 5, 2025Evaluation ResultsMethodMethodLinksBytes per TokenLLaMA-42025.116.05GPT-oss2025.117.35Sutra2025.118.5Sarvam-2B2025.119.42MUTANT-Indic2025.1110.15Gemma-32025.1110.45