Comprehensive Analysis of Transparency and Accessibility of ChatGPT, DeepSeek, And other SoTA Large Language Models

TL;DR

This study analyzes transparency and accessibility of SoTA large language models by framing openness through Open Source Initiative standards and broader definitions of transparency. It consistently contrasts open-source versus open-weight approaches across 112 models released between 2019 and 2025, revealing a predominance of open-weight practices where weights are public but training data, code, and development methodologies remain undisclosed. The findings highlight significant reproducibility, bias mitigation, and accountability limitations associated with partial openness, and they discuss geopolitical and economic trends shaping AI openness. The paper concludes with a call for clearer standards, improved data governance, and sustainable, accountable AI development to ensure responsible deployment of high-performance LLMs.

Abstract

Despite increasing discussions on open-source Artificial Intelligence (AI), existing research lacks a discussion on the transparency and accessibility of state-of-the-art (SoTA) Large Language Models (LLMs). The Open Source Initiative (OSI) has recently released its first formal definition of open-source software. This definition, when combined with standard dictionary definitions and the sparse published literature, provide an initial framework to support broader accessibility to AI models such as LLMs, but more work is essential to capture the unique dynamics of openness in AI. In addition, concerns about open-washing, where models claim openness but lack full transparency, has been raised, which limits the reproducibility, bias mitigation, and domain adaptation of these models. In this context, our study critically analyzes SoTA LLMs from the last five years, including ChatGPT, DeepSeek, LLaMA, and others, to assess their adherence to transparency standards and the implications of partial openness. Specifically, we examine transparency and accessibility from two perspectives: open-source vs. open-weight models. Our findings reveal that while some models are labeled as open-source, this does not necessarily mean they are fully open-sourced. Even in the best cases, open-source models often do not report model training data, and code as well as key metrics, such as weight accessibility, and carbon emissions. To the best of our knowledge, this is the first study that systematically examines the transparency and accessibility of over 100 different SoTA LLMs through the dual lens of open-source and open-weight models. The findings open avenues for further research and call for responsible and sustainable AI practices to ensure greater transparency, accountability, and ethical deployment of these models.(DeepSeek transparency, ChatGPT accessibility, open source, DeepSeek open source)

Figures (6)

• Figure 1: Analysis of NLP/LLM Interest
• Figure 2: Overview of the methodologies used in evaluating ChatGPT, DeepSeek, and SoTA multimodal LLMs
• Figure 3: Illustration of an integrative mind map strategy developed to systematically evaluate transparency and accessibility of 112 LLMs from 2019 to the present. The diagram organizes critical dimensions, including factors influencing model classification, impacts of training methodologies, and consequences of limited data sharing—to comprehensively assess operational efficiency, scalability, and reproducibility challenges in open-weight versus open-source models.
• Figure 4: OSI's first official release of the open source definition, which sets foundational criteria/attributes for Openness in AI
• Figure 5: Temporal distribution of 112 research papers analyzed in this study, spanning from 2019 to 2025. The plot reveals a steadily increasing trend in LLM studies, underscoring rapid advancements in transparency and accessibility.