Back to blog

Blog

OdiaGenAI Releases Llama-3 Fine-Tuned for Odia Language

@openodia
modelsllmresearch

What happened

OdiaGenAI and community contributors released Llama3_Odia — a fine-tuned version of Meta's Llama 3 specifically adapted for the Odia language. This marked the first time a major open-source LLM was optimized for Odia text generation and comprehension.

Why it matters

Large language models are the backbone of modern AI applications — chatbots, content generation, summarization, translation. Having an Odia-specific LLM means developers can build applications that truly understand and produce natural Odia text, rather than relying on English-centric models that treat Odia as an afterthought.

How to use

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("shantipriyap/Llama3_Odia")
tokenizer = AutoTokenizer.from_pretrained("shantipriyap/Llama3_Odia")

inputs = tokenizer("ଓଡ଼ିଆ ଭାଷା ହେଉଛି", return_tensors="pt")
outputs = model.generate(**inputs, max_length=100)
print(tokenizer.decode(outputs[0]))

Links