ResearchBenchmarksTokenization Efficiency on Gujarati (gu)Follow4.69Bytes per TokenLLaMA-44.51965.66986.827.9702Nov 5, 2025Evaluation ResultsMethodMethodLinksBytes per TokenLLaMA-42025.114.69Gemma-32025.116.31GPT-oss2025.116.68Sutra2025.117.36Sarvam-2B2025.117.62MUTANT-Indic2025.118.95