Government Of India
A-
A
A+
ORGANISATION 
Upvote 0
Param-1-Instruct
BharatGen introduces the early checkpoint of SFT (Supervised Fine-Tuned) for Param 1, a bilingual language model trained from scratch in English and Hindi. With 2.9 billion parameters, this checkpoint builds upon the pretraining phase and serves as a foundation for more downstream tasks, safety testing, and customization.
- 0upvotes
19downloads
199views
5.36 GB
0Shares
updated 08 May 2026
Upvote 0
LICENSE Attribution-Non-Commercial 4.0 International (CC BY-NC 4.0)
SECTOR Other
About Model
Pre-Training Details: * Dataset: 7.5 Trillion tokens * Data Quality: Highly curated with standard filtering and multiple processing steps. * Scheduler: Cosine Annealing * Learning_rate: 3e-4 to 3e-6 * Training Setup: Running on 512 H100 GPUs * Framework: NVIDIA NeMo * Precision: bf16-mixed * Base Pre-Trained Checkpoint (Param 1): https://aikosh.indiaai.gov.in/home/models/details/bharatgen_param_1_indic_scale_bilingual_foundation_model.html SFT Training Details: * Dataset: 0.8 Million samples * Epochs: 3 * Scheduler: Cosine Annealing * Learning Rate: 5e-6 to 5e-8 * Training Hardware: 32 H200 GPUs * Framework: NVIDIA NeMo * Precision: bf16-mixed
Metadata
Attribution-Non-Commercial 4.0 International (CC BY-NC 4.0)
Other
bharatgenai
N.A.
N.A.
N.A.
N.A.
N.A.
N.A.
N.A.
N.A.
Open
07/05/26 10:02:42
N.A.
5.36 GB
Tags
Related Models
LayoutXLM - Multilingual Document Understanding
A multimodal pre-trained Transformer model designed for multilingual document AI, integrating text, layout, and image data to improve visually rich document understanding across multiple languages.
LayoutLMv2 - Multimodal Document Understanding
An improved multimodal Transformer model for document AI, integrating text, layout, and image pre-training for enhanced document understanding tasks such as form recognition, document classification, and visual question answering.
LayoutLMv3 - Multimodal Document AI
A pre-trained multimodal Transformer model for document AI, integrating unified text and image masking for tasks such as form understanding, receipt processing, and document layout analysis.
RAG Engine
Python script that implements a Retrieval‑Augmented Generation (RAG) pipeline for document question answering.
tokenmodel
How people read your model before they download it.
Joel-Speech-TTS
A Text-to-Speech (TTS) model developed for the Mizo language that converts written Mizo text into natural-sounding speech, supporting the development of Mizo-language voice applications and accessibility solutions.
mizo-f5-tts-v1
A Text-to-Speech (TTS) model developed for the Mizo language that converts written Mizo text into natural-sounding speech, supporting the development of Mizo-language voice applications and accessibility solutions.
CoRover-ASR-hindi-conformer-hybrid-large
ASR-Hindi-Conformer-Hybrid-Large is a Hindi automatic speech recognition model developed by CoRover, based on the FastConformer architecture with hybrid RNNT and CTC decoding for accurate speech-to-text transcription.
CoRover-BharatGPT-Instruct
BharatGPT-Instruct E2B is a multilingual language model by CoRover.ai, designed to deliver high-quality, culturally relevant AI responses across Indian languages.
CoRover-BharatGPT-mini
BharatGPT Mini - A compact, Transformer-based language model developed by CoRover, pretrained on large-scale text data to understand and generate natural language, advancing India’s indigenous AI capabilities.
Related Models by Organization
sooktam2
Sooktam-2 is a multilingual Indic Text-to-Speech model by BharatGen supporting 12 languages including Hindi, Marathi, Tamil, Telugu, Bengali, Urdu, Punjabi and Indian English. It enables high-quality speech synthesis with reference-guided voice conditioning, preserving speaker voice, accent and prosody for natural and expressive generation.
sooktam2
Sooktam-2 is a multilingual Indic Text-to-Speech model by BharatGen supporting 12 languages including Hindi, Marathi, Tamil, Telugu, Bengali, Urdu, Punjabi and Indian English. It enables high-quality speech synthesis with reference-guided voice conditioning, preserving speaker voice, accent and prosody for natural and expressive generation.
Shrutam-2
Shrutam-2 is a LLM based automatic speech recognition system for 12 major Indian languages. It bridges a Conformer speech encoder with a pretrained LLM decoder through a Mixture-of-Experts (MoE) projection layer, enabling high-quality, prompt-controllable transcription across diverse Indic languages.
Param-1-5B
Param-1-5B is a bilingual (English–Hindi) large language model developed under the Param-1 family. With 5 billion parameters, this model extends the capabilities of Param-1-2.9B by incorporating enhanced mathematical reasoning and code understanding/generation.
The model is pretrained from scratch and designed to serve as a strong foundation for downstream tasks such as mathematical problem solving, and code-related understanding / generation.
BharatGen - Param 1 Indic-Scale Bilingual Foundation Model
Param1 is a 2.9 billion parameter language model pretrained on English and Hindi, designed for text completion.
Param2-17B-Thinking
BharatGen presents Param-2-17B-MoE-A2.4B, a large-scale Mixture-of-Experts (MoE) language model designed to deliver high model capacity while retaining the inference efficiency of a much smaller dense model. It uses a Hybrid MoE architecture with 17B total parameters, while activating only 2.4B parameters per token.
BharatGen Multilingual TTS - Sooktam2
Sooktam-2 is a multilingual Indic Text-to-Speech model by BharatGen supporting 12 languages including Hindi, Marathi, Tamil, Telugu, Bengali, Urdu, Punjabi and Indian English. It enables high-quality speech synthesis with reference-guided voice conditioning, preserving speaker voice, accent and prosody for natural and expressive generation.
BharatGen - Param-1-7B-MoE Advancing Multilingual GenAI for India
Param-1-7B-MoE is a multilingual large language model developed under the Param-1 family as part of BharatGen – A Suite of Generative AI Technologies for India. With 7 billion parameters and a Mixture of Experts (MoE) architecture, the model is designed to better understand and generate text across English, Hindi, and 14 additional Indian languages.
The model is pretrained from scratch with a strong focus on linguistic diversity, cultural context, and large-scale multilingual representation.
BharatGen-AgriParam
Large language model fine-tuned from Param-1-2.9B-Instruct on a high-quality, India-centric agriculture dataset.
BharatGen-FinanceParam
large language model fine-tuned from Param-1-2.9B-Instruct on a high-quality finance dataset.
Related Models by Sector
RAG Engine
Python script that implements a Retrieval‑Augmented Generation (RAG) pipeline for document question answering.
Joel-Speech-TTS
A Text-to-Speech (TTS) model developed for the Mizo language that converts written Mizo text into natural-sounding speech, supporting the development of Mizo-language voice applications and accessibility solutions.
mizo-f5-tts-v1
A Text-to-Speech (TTS) model developed for the Mizo language that converts written Mizo text into natural-sounding speech, supporting the development of Mizo-language voice applications and accessibility solutions.
CoRover-ASR-hindi-conformer-hybrid-large
ASR-Hindi-Conformer-Hybrid-Large is a Hindi automatic speech recognition model developed by CoRover, based on the FastConformer architecture with hybrid RNNT and CTC decoding for accurate speech-to-text transcription.
CoRover-BharatGPT-Instruct
BharatGPT-Instruct E2B is a multilingual language model by CoRover.ai, designed to deliver high-quality, culturally relevant AI responses across Indian languages.
CoRover-BharatGPT-mini
BharatGPT Mini - A compact, Transformer-based language model developed by CoRover, pretrained on large-scale text data to understand and generate natural language, advancing India’s indigenous AI capabilities.
CoRover-BharatGPT-3B-Indic
A multilingual model fine-tuned on authentic conversational data across Indic languages, supporting translation, summarisation, and text generation. Ideal for Secure RAG (S-RAG) and domain-specific fine-tuning, it enables the development of sovereign, multilingual AI agents, chatbots, voicebots, and virtual assistants.
Streetlight Working or Not-Working Classification
MobileNetV2-based binary image classification model for automated streetlight visual inspection built using the Aargus DIY visual inspection tool. The model classifies streetlights as Working or Not Working from images or video frames and is intended to support municipal infrastructure monitoring, visual inspection, smart-city applications and maintenance prioritization. The model achieves 95.16% accuracy, 91.30% precision and 95.45% recall on the reported evaluation data.
Welding Defect Detection
A YOLOv8-based computer vision model developed using the Aargus DIY Visual Inspection Tool for automated visual inspection of welds. The model detects and classifies three categories: Bad Weld, Good Weld, and Defect. It is designed to support automated industrial quality inspection and welding defect identification from images.
MIZO LANGUAGE OCR
This dataset is designed for training and evaluating Optical Character Recognition (OCR) models capable of extracting text from document images. It contains 14,000 labeled samples consisting of image filenames and their corresponding ground-truth text annotations, along with metadata such as font type, font size, and document category. The dataset includes diverse document styles, including printed books and scanned documents, enabling models to learn robust text recognition under varying visual
Related Models by Task Type
LayoutXLM - Multilingual Document Understanding
A multimodal pre-trained Transformer model designed for multilingual document AI, integrating text, layout, and image data to improve visually rich document understanding across multiple languages.
LayoutLMv2 - Multimodal Document Understanding
An improved multimodal Transformer model for document AI, integrating text, layout, and image pre-training for enhanced document understanding tasks such as form recognition, document classification, and visual question answering.
LayoutLMv3 - Multimodal Document AI
A pre-trained multimodal Transformer model for document AI, integrating unified text and image masking for tasks such as form understanding, receipt processing, and document layout analysis.
tokenmodel
How people read your model before they download it.
Gemma-4-12B-it-assistant
This model card is for the Multi-Token Prediction (MTP) drafters for the Gemma 4 models. MTP is implemented by extending the base model with a smaller, faster draft model. When used in a Speculative Decoding pipeline, the draft model predicts several tokens ahead, which the target model then verifies in parallel. This results in significant decoding speedups (up to 3x) while guaranteeing the exact same quality as standard generation, making these checkpoints perfect for low-latency and ODML.
DiffusionGemma
DiffusionGemma is a generative model built by Google DeepMind. Based on the 26B A4B Mixture-of-Experts (MoE) Gemma 4 architecture, DiffusionGemma generates tokens using discrete diffusion. This open-weights model is multimodal, handling text, image, and video inputs to generate text output.
Med-Doc-Classifier
A document/image classifier for indian healthcare setting
Parrotlet-V-2p5-pro
Model for schema-driven extraction from Indian medical records: prescriptions, invoices, lab reports, and discharge summaries, plus PII detection.
Document-PII-Redactor
Image PII redaction tuned for indian medical documents
NEET-BioBERT
NEET-BioBERT is a fine-tuned version of DistilBERT (base uncased) specifically trained to classify the correct option for NEET-style multiple-choice biology questions. It selects the best answer among four choices (A, B, C, D).
© 2026 - Copyright AIKosh. All rights reserved.