A recent quality audit of OpenDeepWiki, a code knowledge base generated using GPT-5.5 for directory structure and Mimo-2.5 for content, reveals a groundbreaking cost-efficiency milestone. The project consumed approximately 1.3 billion tokens at a total cost of just 39 RMB (roughly $5.40), translating to about 0.03 RMB per million tokens. This extreme low-cost approach challenges assumptions about the resources required to build comprehensive code documentation. The audit examined sample pages from both OpenDeepWiki and AIDotNet/Means, evaluating the quality of generated content. While the results show promising potential for democratizing knowledge base creation, they also highlight areas where generated content may require human oversight. For engineering teams and indie hackers, this signals a new possibility: building custom, large-scale code knowledge bases without prohibitive cloud costs.
A quality audit of OpenDeepWiki reveals how 1.3B tokens of code knowledge were generated for just 39 RMB using GPT-5.5 and Mimo-2.5.