This empirical study offers a data-driven perspective on how LLMs are actually being used, highlighting several themes that nuance the conventional wisdom about AI deployment:
1. A Multi-Model Ecosystem. Our analysis shows that no single model dominates all usage. Instead, we observe a rich multi-model ecosystem with both closed and open models capturing significant shares. For example, even though OpenAI and Anthropic models lead in many programming and knowledge tasks, open source models like DeepSeek and Qwen collectively served a large portion of total tokens (sometimes over 30%). This suggests the future of LLM usage is likely model-agnostic and heterogeneous. For developers, this means maintaining flexibility, integrating multiple models and choosing the best for each job, rather than betting everything on one model's supremacy. For model providers, it underscores that competition can come from unexpected places (e.g., a community model might erode part of your market unless you continuously improve and differentiate).
2. Usage Diversity Beyond Productivity. A surprising finding is the sheer volume of roleplay and entertainment-oriented usage. Over half of open source model usage was for roleplay and storytelling. Even on proprietary platforms, a non-trivial fraction of early ChatGPT use was casual and creative before professional use cases grew. This counters an assumption that LLMs are mostly used for writing code, emails, or summaries. In reality, many users engage with these models for companionship or exploration. This has important implications. It highlights a substantial opportunity for consumer-facing applications that merge narrative design, emotional engagement, and interactivity. It suggests new frontiers for personalization—agents that evolve personalities, remember preferences, or sustain long-form interactions. It also redefines model evaluation metrics: success may depend less on factual accuracy and more on consistency, coherence, and the ability to sustain engaging dialog. Finally, it opens a pathway for crossovers between AI and entertainment IP, with potential in interactive storytelling, gaming, and creator-driven virtual characters.
3. Agents vs Humans: The Rise of Agentic Inference. LLM usage is shifting from single-turn interactions to agentic inference, where models plan, reason, and execute across multiple steps. Rather than producing one-off responses, they now coordinate tool calls, access external data, and iteratively refine outputs to achieve a goal. Early evidence shows rising multi-step queries and chained tool use that we proxy to agentic use. As this paradigm expands, evaluation will move from language quality to task completion and efficiency. The next competitive frontier is how effectively models can perform sustained reasoning, a shift that may ultimately redefine what agentic inference at scale means in practice.
4. Geographic Outlook. LLM usage is becoming increasingly global and decentralized, with rapid growth beyond North America. Asia's share of total token demand has risen from about 13% to 31%, reflecting stronger enterprise adoption and innovation. Meanwhile, China has emerged as a major force, not only through domestic consumption but also by producing globally competitive models. The broader takeaway: LLMs must be globally useful performing well across languages, contexts, and markets. The next phase of competition will hinge on cultural adaptability and multilingual capability, not just model scale.
5. Cost vs. Usage Dynamics. The LLM market does not seem to behave like a commodity just yet: price alone explains little about usage. Users balance cost with reasoning quality, reliability, and breadth of capability. Closed models continue to capture high-value, revenue-linked workloads, while open models dominate lower-cost and high-volume tasks. This creates a dynamic equilibrium—one defined less by stability and more by constant pressure from below. Open source models continuously push the efficient frontier, especially in reasoning and coding domains (e.g. Kimi K2 Thinking) where rapid iteration and OSS innovations narrow the performance gap. Each improvement in open models compresses the pricing power of proprietary systems, forcing them to justify premiums through superior integration, consistency, and enterprise support. The resulting competition is fast-moving, asymmetric, and continuously shifting. Over time, as quality convergence accelerates, price elasticity is likely to increase, turning what was once a differentiated market into a more fluid one.
6. Retention and the Cinderella Glass Slipper Phenomenon. As foundation models advance in leaps, not steps, retention has become the true measure of defensibility. Each breakthrough creates a fleeting launch window where a model can "fit" a high-value workload perfectly (the Cinderella Glass Slipper moment) and once users find that fit, they stay. In this paradigm, product-market fit equals workload-model fit: being the first to solve a real pain point drives deep, sticky adoption as users build workflows and habits around that capability. Switching then becomes costly, both technically and behaviorally. For builders and investors, the signal to watch isn't growth but retention curves, namely, the formation of foundational cohorts who stay through model updates. In an increasingly fast-moving market, capturing these important unmet needs early determines who endures after the next capability leap.
Together, LLMs are becoming an essential computational substrate for reasoning-like tasks across domains, from programming to creative writing. As models continue to advance and deployment expands, having accurate insights on real-world usage dynamics will be crucial for making informed decisions. Ways in which people use LLMs do not always align with expectations and vary significantly country by country, state by state, use case by use case. By observing usage at scale, we can ground our understanding of LLM impact in reality, ensuring that subsequent developments, be they technical improvements, product features, or regulations, are aligned with actual usage patterns and needs. We hope this work serves as a foundation for more empirical studies and that it encourages the AI community to continuously measure and learn from real-world usage as we build the next generation of frontier models.