Developer James O'Claire discovered a stark pricing reality while testing Hermes web research with DeepSeek V4, finding nearly a 50x price increase compared to Anthropic and OpenAI's frontier models. The cost differential becomes even more pronounced when accounting for token usage efficiency, where closed models consume significantly more tokens for equivalent tasks.

O'Claire questions whether this pricing strategy stems from open weight models benefiting from community optimization across diverse hardware, or if these models serve as loss leaders to pressure competitors downward. He draws parallels to luxury branding, suggesting Anthropic and OpenAI deliberately gate their frontier models to create artificial scarcity rather than compete on price.

The analysis extends to regulatory concerns, warning that these companies may leverage geopolitical tensions to restrict open weight model access. This follows a pattern where Google Gemma 4, Meta's Llama, and OpenAI's last open releases (dating to 2025) trail behind truly open alternatives.

True open source initiatives like Allen AI's OLMO models are gaining traction despite data cutoff limitations (December 2024). A significant NSF and Nvidia partnership now enables fully open AI development, representing the most viable path forward for accessible artificial intelligence without artificial constraints.