Micah Hill-Smith is co-founder and chief executive of Artificial Analysis, the independent AI benchmarking company he founded with George Cameron. Its work helps engineers select models and infrastructure by measuring capabilities alongside speed, cost, availability, and performance on practical tasks.
Hill-Smith studied law and computer science at the University of Auckland, experimented with GPT-2, and met Cameron during an internship at Google. From 2020 to 2022, he worked as a business analyst in McKinsey’s Sydney office, including on strategy projects for major banks.
While developing an AI legal-research tool, Hill-Smith discovered that different stages required different language models, but dependable comparisons of their quality, speed, and price were scarce. He began building comparison dashboards in early 2023 before joining Cameron to establish Artificial Analysis. He later described the company’s origins as an extension of that model-selection problem.
Artificial Analysis expanded from language-model comparisons into agents, multimodal systems, inference providers, and AI accelerators. Its State of AI reports combine capability evaluations, performance measurements, pricing, and user-preference data.
- Multidimensional model evaluation. The Artificial Analysis Intelligence Index compares general model capabilities, while accompanying measurements account for inference speed, provider differences, price, and access to open-weight models. Together, they help builders identify systems suited to specific applications instead of assuming the highest-ranked model is always the best choice.
- Task-level inference economics. Reasoning models can consume substantially more output tokens, increasing costs and response times even when advertised per-token prices appear competitive. Those delays compound across sequential agent requests, making token consumption and end-to-end latency essential deployment considerations.
- AA-Briefcase. Hill-Smith announced this agent benchmark in June 2026. Its evaluation framework tests realistic professional projects in areas including data science, product management, and corporate strategy, assessing whether deliverables meet requirements while measuring analytical quality, presentation, and cost.
- International model competition. Hill-Smith tracks American and Chinese AI laboratories through comparative rankings and the growing availability of open-weight systems. In an August 2026 assessment, he estimated that leading Chinese models historically trailed leading American systems by several months while increasingly competing near the frontier.
