Krutrim-1 Instruct Large Language Model

About Model

Krutrim-1 is a transformer-based Large Language Model (LLM) developed by OLA Krutrim Labs, optimized for multilingual Indic benchmarks. It has been trained on the largest known Indic-language dataset, addressing challenges in low-resource Indian dialects and linguistic diversity. The model demonstrates strong performance on Indic language benchmarks, matching or surpassing models like LLAMA-2 in multilingual fluency and generation tasks. It supports various NLP applications, including text generation, creative writing, summarization, and translation. Krutrim-1 features 32 transformer layers, 48 attention heads, and a vocabulary size of 70,400. Released in January 2024, it is available in both base and instruction-tuned versions. Ethical safeguards have been implemented, but users are advised to exercise human oversight, as the model may reflect biases from its training data.