[Open Source vs. Closed Source 'Money Grab Battle': Price Gap Reaches 40x, Open Source Models Diverting Billions in Revenue from Leading Giants] According to monitoring by 动察 Beating, following the concentrated release of 25 open-source weight models such as NVIDIA's Nemotron 3 Ultra and Google's Gemma 4 12B, tech investor Chamath Palihapitiya pointed out that the capability gap between open-source and closed-source frontier models is rapidly narrowing, but there remains a massive disparity in usage costs. Based on calculations for enterprises consuming 1 billion input and 1 billion output tokens per month, GPT-5.5 Pro costs as much as $105,000, Claude Opus 4.8 costs $30,000, while DeepSeek R1 costs only $2,740, with GPT-5.5 Pro carrying a nearly 40x premium. Chamath warned that most company CEOs are unaware that their technical teams, in the absence of governance and auditing, are defaulting to the most expensive large model APIs, leading to budget overruns. As model routing control planes, such as the AI-native software development lifecycle orchestration platform Software Factory, which Chamath co-founded and serves as CEO, become increasingly widespread, enterprises will shift to model-agnostic architectures. Large-scale inference will be defaulted to DeepSeek, high-end proxy flows will be routed to Claude Opus, and GPT-5.5 Pro will only be invoked on-demand when clear, significant incremental value is generated. Chamath predicts that this refined routing approach will significantly slow the API revenue growth of frontier labs like OpenAI and Anthropic, while the revenue of open-source and low-cost inference ecosystems will see explosive growth. [Original Article Link]