What Artificial Analysis does
Artificial Analysis (artificialanalysis.ai) runs its own benchmarks on AI models, the cloud APIs that host them and the chips they run on, and publishes the results as public leaderboards and charts. Its About page lists 500+ models benchmarked, 100+ inference providers and 1,000+ endpoints. The headline number is the Artificial Analysis Intelligence Index, a composite score. Version 5, announced October 27, 2026, weights knowledge work, coding, science and general knowledge at 25% each, adds Cognition's FrontierCode 2.0 and Terminal-Bench-Science, and says 60% of the index now rests on private test sets or held-out answers. Beyond language models it covers coding agents, text-to-image, video, text-to-speech, speech-to-text and music, including community arenas for image, video and speech.
Performance benchmarks measure what customers see end to end (output speed, time to first token, end-to-end time per task) rather than the peak a chip can reach, and prices are tracked per endpoint. The company says providers cannot pay for results, methodology changes or listing. Signed-in tools include a Model Recommender, MicroEvals for comparing models on your own prompts, and Optima for turning your own tasks into a repeatable benchmark that compares quality against cost per task. Data is also available through a REST API and an MCP server.
