Claude Opus 5

ASSESSMENT

Claude Opus 5 sits at the top of the independently verified standings as of the August survey, with an Artificial Analysis intelligence index of 61 — the highest independent mark in the field. Anthropic ships it with extended thinking enabled by default and an optional Fast mode, positioning it as a reasoning-first flagship rather than a general-purpose all-rounder.

At $5 / $25 per million tokens it is priced as a premium model, though its input price is notably half that of Fable 5 — a sign of Anthropic consolidating its top tier around Opus. The roughly 1M-token context suits long documents, large codebases, and multi-step agentic work, which is exactly where the model's lead is most visible.

The costs are real. Output at $25 is among the steepest in the survey, the deepest reasoning modes add latency and token spend, and the weights are closed, so self-hosting is off the table. For teams that need the best available reasoning and agentic reliability and can absorb the price, Opus 5 is the current reference point; for cost-sensitive or high-volume work, its own cheaper siblings or the open-weight tier will often be the better fit.