DeepSeek V4 Pro: The First Chinese Model at the Frontier (foodtruckbench.com)

DeepSeek V4 Pro has emerged as the first Chinese model in the "frontier" ROI tier, achieving a remarkable fourth place on the competitive leaderboard. This model finished its runs with median net worth exceeding $27,000—3.5 times higher than the previous top Chinese model and incredibly consistent, with all five runs falling within a narrow $26K-$35K range. Its performance indicates a significant narrowing of the gap between Chinese and Western AI models, reflecting a maturation of China’s AI capabilities, as it now competes closely with leading models like GPT-5.2 and Opus 4.6 rather than just surviving. The significance of this development lies in the model’s efficient operational strategy, leveraging persistent memory and structured planning akin to that of a seasoned operator. DeepSeek V4 Pro’s skill at minimizing food waste, achieving zero debt, and producing high revenues substantiates its classification as a competitive frontier model. Its unique approach to using a detailed key-value store as a working scratchpad allows it to execute coherent operational plans effectively. With its cost-effectiveness—at $0.88 per run—DeepSeek V4 Pro not only challenges previously established norms of model efficiency but also redefines what constitutes viable options in the premium-tier AI market, setting a new benchmark for future innovations.