ResearchBenchmarksTokenization Efficiency on EnglishFollow3.65Bytes per TokenSarvam-2B3.57844.06174.5455.0283Nov 5, 2025Evaluation ResultsMethodMethodLinksBytes per TokenSarvam-2B2025.113.65Gemma-32025.114.36LLaMA-42025.114.53GPT-oss2025.114.55Sutra2025.115.15MUTANT-Indic2025.115.44