ResearchBenchmarksTokenization Efficiency on Telugu (te)Follow4.43Bytes per TokenLLaMA-44.17925.87217.5659.2579Nov 5, 2025Evaluation ResultsMethodMethodLinksBytes per TokenLLaMA-42025.114.43GPT-oss2025.116.41Gemma-32025.116.82Sutra2025.118.04Sarvam-2B2025.118.46MUTANT-Indic2025.1110.7