Dhruv 27b
Open source AI models built for and by India, with next-generation capabilities and culturally nuanced understanding.
Dhruv-27b
Our flagship generative AI model focused on research, reasoning, and deep understanding of nuanced contexts.
Key Capabilities
- •Complex reasoning across multiple domains including mathematics, science, code, and humanities
- •Deep understanding of Indian contexts, languages, and cultural references
- •Advanced knowledge synthesis from multiple sources with citation capabilities
- •Proprietary Thoughts Layer integration for unparalleled reasoning capabilities
Model Specs
- Parameters:
- 27 billion
- Context:
- 128K input and 8k output
- Architecture:
- Transformer-based with custom attention mechanisms
- Training:
- Curated datasets with focus on Indian content and high-quality research materials
Performance Benchmarks
Dhruv-27b performance compared to other leading models
MMLU (Multi-task Language Understanding)82.4%
GPQA Diamond67.8%
MATH90.8%
Facts Grounding77.8%
Monthly Updates
Dhruv models receive continuous improvements
April 2025
Significant improvements in UI generation, code generation accuracy, and overall reasoning abilities.
March 2025
Enhanced mathematical reasoning capabilities and improved code generation