Artificial Analysis

Artificial Analysis is an independent benchmarking and comparison surface for AI models, API providers, and hardware. It publishes the Artificial Analysis Intelligence Index (currently v4.0, combining 10 evaluations including GDPval-AA, τ²-Bench Telecom, Terminal-Bench Hard, SciCode, AA-LCR, AA-Omniscience, IFBench, Humanity's Last Exam, GPQA Diamond, and CritPt), plus specialised Coding and Agentic indices, image/video/speech leaderboards, an Openness Index, API Provider Performance ranks, and the new AA-AgentPerf hardware benchmark for real agent workloads. Used by buyers choosing models/providers and by the industry as a shared scoreboard. Free public site with changelog-driven evaluation updates.