Artificial intelligence has been hailed as a game-changer for productivity, with promises of revolutionizing how we work. However, a recent study by OpenRouter has uncovered a different reality by analyzing over 100 trillion tokens, revealing surprising insights into real-world AI usage. The study provides a unique glimpse into how AI is deployed globally, challenging preconceived notions about its impact.
One of the most unexpected findings from the study is that over half of open-source AI model interactions are not related to productivity but rather focus on roleplay and creative storytelling. Users are engaging in character-driven conversations and interactive fiction, highlighting a significant shift in how AI is being utilized. This trend showcases a growing interest in using AI for companionship and exploration rather than strictly for work-related tasks.
Furthermore, the study highlights the meteoric rise of programming-related queries across all AI models. Coding queries have surged from 11% to over 50% of total AI usage, underscoring the deep integration of AI into software development. Developers are increasingly relying on AI models for complex debugging sessions and architectural reviews, reshaping the landscape of AI usage in the programming realm.
Moreover, the study reveals a significant surge in Chinese AI models, which now account for 30% of global AI usage, challenging the dominance of Western companies. Models from DeepSeek, Qwen (Alibaba), and Moonshot AI have gained traction, signaling a shift in the global AI landscape. Additionally, the concept of agentic inference emerges as a defining trend, showcasing AI models’ evolution from text generators to autonomous agents capable of multi-step tasks and reasoning. This shift signifies a new phase in AI development, with models becoming more adept at planning and execution.
