The video starts with platform data showing open-weight modelAn open-weight AI model makes its learned parameter values available for others to download, inspect or run under a stated license. usage rising while closed-modelA closed-weight AI model keeps its learned weights unavailable and is commonly accessed through a provider-controlled service. share falls. DeepSeek is presented as an example: it can rival Anthropic's token share while representing only a small fraction of total spend because its tokens are dramatically cheaperToken pricing is the rate an AI provider charges for processing input tokens, generating output tokens, or reading cached tokens..
That gap creates two complementary markets. Open models expand total demand, give users control over deployment and data, and let companies fine-tune models for specific domainsDomain-specific AI model fine-tuning adapts a pretrained model using examples from a particular field, workflow or organization.. Frontier providers retain pricing power when a small quality advantage changes the outcome of a difficult, high-value task.
The host argues that businesses should evaluate cost per completed taskCost per completed task measures the total AI, tool, infrastructure, retry, and repair expense for each verified useful outcome. rather than headline token prices. A cheaper model may need more tokens or retries to finish the same work, while a specialized open model can become far more valuable after training on private enterprise knowledge.
Open weights also reduce platform risk by giving companies a choice of self-hosting or competing inference providers. The video concludes that open models are likely to win volume while OpenAI and Anthropic keep much of the revenue, with a separate concern that US companies may become dependent on Chinese model and chip ecosystems. The Higgsfield sponsorship segment is omitted from the analysis.
Watch on YouTube
