Anthropic builds the Claude family and holds the current independent lead with Claude Opus 5. Its models anchor the flagship range of the survey; safety and interpretability research are central to how it ships.
Anthropic holds the independent lead of the survey. Claude Opus 5 tops the register at an Artificial Analysis index of 61, the highest independently verified mark recorded here, and the lab fields the widest tier coverage of any single party — Opus 5 and Fable 5 at flagship, Sonnet 5 at balanced, Haiku 4.5 at efficient. No other maker spans all three closed tiers.
Safety and interpretability research sit at the center of how the lab ships, and that posture reads in a release cadence that favors measured, well-documented launches over rapid iteration. Weights are closed across the entire range; there is no open-weight Claude, which cedes the self-hosting and fine-tuning market to the open country by design.
Pricing runs at the premium end of the field, the flagship tier's clearest cost-for-capability tradeoff — buyers paying for Opus are paying for the top independent standing, not the lowest rate. The lead is real but narrow: GPT-5.6 Sol trails by only two index points, and the gap at the frontier is measured in single digits.