BharatGPT-3B-Indic

Model Description

This model is fine-tuned and designed to generate multilingual outputs across multiple Indic languages. The model has been trained on a diverse and curated dataset comprising Hindi, Punjabi, Marathi, Malayalam, Oriya, Kannada, Gujarati, Bengali, Urdu, Tamil, and Telugu. It is optimized to handle natural language tasks such as translation, summarization, and conversational generation across these languages. This model is trained on authentic Indian conversational data in 12 languages. However, it is not designed for direct use as a standalone chatbot, as it lacks the latest data updates. It is best suited for S-RAG (Secure Retrieval-Augmented Generation) or fine-tuning with your own data. For enhanced performance, integration with Conversational Agentic AI Platform is recommended (though not mandatory). This platform enables the creation of multi-modal and multi-lingual AI Agents, Co-Pilots, and Virtual Assistants (such as ChatBots, VoiceBots, and VideoBots) using a sovereign AI and composite AI approach. It leverages classic NLP, grounded generative AI, and Generally Available LLMs to deliver powerful, versatile solutions.

• Developed by: CoRover.ai
• Model type: Finetuned (Language Model for Multilingual Text Generation and Text Completion)
• Language(s) (NLP): Hindi, Punjabi, Marathi, Malayalam, Oriya, Kannada, Gujarati, Bengali, Urdu, Tamil, Telugu
• Learn (Become C-CAP: CoRover Certified AI Professional): Get Certified in 1 Hour

How to Get Started with the Model

Make sure to update your transformers and bitsandbytes installation via pip install -U transformers & pip install -U bitsandbytes

Use the code below to get started with the model.

English

import torch
from transformers import pipeline

model_id = "CoRover/BharatGPT-3B-Indic"
pipe = pipeline(
    "text-generation",
    model=model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
    trust_remote_code=True
)
messages = [
    {"role": "system", "content": "You are a helpful assistant who responds in English"},
    {"role": "user", "content": "who created you?"},
]
outputs = pipe(
    messages,
    max_new_tokens=256,
)
print(outputs[0]["generated_text"][-1])

Hindi

import torch
from transformers import pipeline

model_id = "CoRover/BharatGPT-3B-Indic"
pipe = pipeline(
    "text-generation",
    model=model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
    trust_remote_code=True
)
messages = [
    {"role": "system", "content": "You are a helpful assistant who responds in Hindi"},
    {"role": "user", "content": "भारत की राजधानी क्या है"},
]
outputs = pipe(
    messages,
    max_new_tokens=256,
)
print(outputs[0]["generated_text"][-1])

Gujarati

import torch
from transformers import pipeline

model_id = "CoRover/BharatGPT-3B-Indic"
pipe = pipeline(
    "text-generation",
    model=model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
    trust_remote_code=True
)
messages = [
    {"role": "system", "content": "You are a helpful assistant who responds in Gujarati"},
    {"role": "user", "content": "શું છે ભારતની રાજધાની"},
]
outputs = pipe(
    messages,
    max_new_tokens=256,
)
print(outputs[0]["generated_text"][-1])

Marathi

import torch
from transformers import pipeline

model_id = "CoRover/BharatGPT-3B-Indic"
pipe = pipeline(
    "text-generation",
    model=model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
    trust_remote_code=True
)
messages = [
    {"role": "system", "content": "You are a helpful assistant who responds in Marathi"},
    {"role": "user", "content": "भारताची राजधानी कोणती?"},
]
outputs = pipe(
    messages,
    max_new_tokens=256,
)
print(outputs[0]["generated_text"][-1])

Malayalam

import torch
from transformers import pipeline

model_id = "CoRover/BharatGPT-3B-Indic"
pipe = pipeline(
    "text-generation",
    model=model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
    trust_remote_code=True
)
messages = [
    {"role": "system", "content": "You are a helpful assistant who responds in Malayalam"},
    {"role": "user", "content": "ഭരത് കി രാജധാനി ഉണ്ട്"},
]
outputs = pipe(
    messages,
    max_new_tokens=256,
)
print(outputs[0]["generated_text"][-1])

Multilingual MMLU Benchmarks

BharatGPT-Instruct evaluated across Indian languages against Qwen-1.7B and Sarvam-30B.

RAG Evaluation

This is the most relevant comparison for our actual product surface-enterprise AI-Assistant grounded in a customer knowledge base. BharatGPT-Instruct scores 100.00 on Faithfulness, essentially tied with GPT-4o-mini and Qwen 3 1.7B, and matches both on Top-K Accuracy (%).

Benchmark Results

BharatGPT models were evaluated on a diverse and comprehensive set of benchmark datasets and evaluation metrics to measure their performance across multiple dimensions of language understanding, reasoning, and text generation. The reported results cover a broad range of tasks, including knowledge-based question answering, reasoning, comprehension, and safty and Truthfulness.

Training Details

Training Data

• Language Spectrum: A balanced representation of Hindi, Punjabi, Marathi, Malayalam, Oriya, Kannada, Gujarati, Bengali, Urdu, Tamil, and Telugu, capturing the unique syntactic structures of each language.

Usage and Limitations

• License: Non-Commercial. For academic and research purposes only. For commercial use, please visit Conversational Agentic AI Platform or Contact Us.

• Terms of Use: Terms and Conditions

• PIPEE -Responsible AI Framework:

  • P - Purpose: Risk depends on context.
  • I - Inclusivity: As the Hon'ble PM of India says, "AI means All Inclusive" - it must be human-centric and multi-modal - Text, Voice, and Video.
  • P - Privacy: Where data is stored and inferred.
  • E - Explainability: Can we audit the decision?
  • E - Energy Efficiency: Just because someone has money for GPUs doesn't mean we should waste energy.

Hardware & Software

To ensure top-tier performance and scalability, the model was fine-tuned using state-of-the-art hardware and software configurations:

• NVIDIA A100 GPUs, renowned for their unmatched computational power and efficiency in deep learning tasks, were leveraged. These GPUs, with their advanced tensor cores, provided the ability to train large-scale models with reduced training time and enhanced precision. High-bandwidth GPU interconnects ensured seamless parallel processing for handling massive multilingual datasets.