ResearchBenchmarksTokenization Efficiency on Malayalam mlFollow5.26Bytes per TokenLLaMA-45.01726.65618.2959.9339Nov 5, 2025Evaluation ResultsMethodMethodLinksBytes per TokenLLaMA-42025.115.26GPT-oss2025.117.43Gemma-32025.117.57Sarvam-2B2025.118.63Sutra2025.118.75MUTANT-Indic2025.1111.33