Artificial Analysis is an independent AI benchmarking service that compares models and API hosting providers on quality, price, output speed and latency. Its Intelligence Index combines 10 evaluations, including Terminal-Bench 4.0, SciCode and Humanity's Last Exam, and 690 models are tracked across language, image, video and speech. Charts show intelligence against cost per task, open weights against proprietary models and progress over time. A Model Recommender and custom benchmark builder help match models to priorities. The Pro plan costs $417 per seat per month and adds API access, data export, custom charts and email support, while Enterprise pricing is custom.
The service is built for developers, AI engineers, product managers, researchers and analysts who must pick a model or inference provider and want numbers rather than marketing claims. Small organizations can use Pro for data exports and reports, while larger companies with 150 or more employees can go to Enterprise for the highest API rate limits, workshops, AI strategy advisory and personalized support. Investors and strategists also read its reports to follow how labs and countries compare. Anyone weighing open weights models against closed ones will find the side-by-side comparisons useful.
An engineering lead choosing a model for a coding agent can check the Coding Agent Index and the cost per task before committing. A startup comparing hosts for the same open model can look at output tokens per second and latency across providers. Security teams can follow the Cyber Index, and media teams can consult the text to image leaderboard or speech preference ratings. Analysts download data to build custom tables and charts, and consultants draw on reports such as the State of AI report and the AI Adoption Survey. The changelog records new model evaluations as they are published, keeping comparisons current.
