Artificial Analysis Launches Six New AI Capability Indices for Professional Domains

Artificial Analysis, the independent AI benchmarking platform, just rolled out six new Capability Indices. Each one evaluates frontier AI models against real-world tasks in a specific professional domain: finance and accounting, legal, healthcare and medical, strategy and ops, engineering, and economics.

The new indices are composite benchmarks, meaning they blend multiple evaluation metrics into a single score per domain. They incorporate the platform’s GDPval-AA v2 framework, which tests AI-generated deliverables across 44 occupations and 9 industries relevant to these six sectors.

The Legal Index offers a useful window into how the weighting works. Legal Knowledge accounts for 35% of the score, Agentic Knowledge Work makes up 25%, and Reasoning contributes 15%. The remaining share is distributed across other evaluative criteria.

Early results show variants of Claude Fable 5 and Claude Opus 4.8 sitting near the top of the Legal Index leaderboard. No specific numerical scores have been published yet, but the relative rankings are visible on the platform’s site.

Artificial Analysis’s Intelligence Index, now at version 4.1, was already tilting toward agentic and practical task evaluation. The six new Capability Indices extend that philosophy into vertical-specific territory.

The GDPval-AA v2 framework, which underpins the new indices, tries to answer that kind of question by simulating the actual deliverables professionals produce. A finance benchmark that tests whether an AI can generate a credible DCF model is fundamentally different from one that tests whether it can define “discounted cash flow.”

For readers wondering what this means for crypto markets: this particular development doesn’t have a direct line to token prices or blockchain protocols. There are no tokenized benchmark scores, no on-chain evaluation frameworks, and Artificial Analysis has made no mention of crypto assets in connection with these indices.