ResearchBenchmarksTokenization Efficiency on Tamil taFollow4.1Bytes per TokenLLaMA-43.81125.76067.719.6594Nov 5, 2025Evaluation ResultsMethodMethodLinksBytes per TokenLLaMA-42025.114.1GPT-oss2025.117.56Sutra2025.119.38Gemma-32025.119.6Sarvam-2B2025.119.74MUTANT-Indic2025.1111.32