Deepseek Open-Source AI

Tripti R, Kulkarni and Chethan, P and Arun Vikas, Singh and Monisha S., S (2025) Deepseek Open-Source AI. International Journal of Trend in Scientific Research and Development, 9 (3). pp. 555-560. ISSN 2456-6470

Text ijtsrd80053.pdf Download (945kB)

Abstract

DeepSeek AI represents a new paradigm in the evolution of large-scale language and vision models, integrating advanced machine learning techniques with real-time multimodal capabilities. This research explores the core architecture, training methodology, and performance benchmarks of DeepSeek-VL and DeepSeek-Coder, opensource models capable of state-of-the-art reasoning in text, image, and code generation. With the growing demand for transparent and scalable AI, DeepSeek presents a significant leap toward democratizing access to high performing foundational models. Results indicate superior performance on established benchmarks such as VQAv2, COCO Captioning, and MMLU. This paper offers an analytical perspective on the architecture and training pipeline, emphasizing its relevance to academic and industrial applications.

Item Type:	Article
Subjects:	Q Science > Q Science (General)
Divisions:	Postgraduate > Master's of Islamic Education
Depositing User:	Journal Editor
Date Deposited:	26 May 2025 10:53
Last Modified:	27 May 2025 12:58
URI:	http://eprints.umsida.ac.id/id/eprint/16117

Actions (login required)

View Item