Anthropic has released Claude Opus 5, and it went straight to the top. On launch day it ranks number one on both of Artificial Analysis's independent leaderboards, the Intelligence Index and the Agentic Index, ahead of Anthropic's own Fable 5 and OpenAI's GPT-5.6 Sol.
This is Anthropic's fourth model in under two months, after Mythos 5, Fable 5, and Sonnet 5, a cadence that tells you as much about the market as the model does. What makes Opus 5 notable isn't just that it leads, but that it leads while costing the same $5 / $25 per million tokens as Opus 4.8, roughly half the price of Fable 5 for near-equal performance.
What actually changed
Two things stand out for real work. First, a new effort setting (low to high, plus a max tier) lets you dial intelligence against token cost per request, pay for deep reasoning only when the task needs it. Second, the gains are largest exactly where they matter: on Frontier-Bench it more than doubles Opus 4.8's score, and on agentic and long-horizon coding tasks it's within a hair of Fable 5's peak at half the cost per task. There's also a fast mode that runs about 2.5× quicker for double the price when latency matters more than budget.
What it means for you
For anyone using Claude, or building on it, this is the kind of upgrade that quietly resets the price-to-capability line for the whole market. A model that tops the independent leaderboards at half the previous flagship's cost pressures every competitor's pricing, and makes agentic features that were borderline-affordable last month viable this month. If you picked a model even a few weeks ago, it's worth re-checking your options against ChatGPT and the rest, the frontier moved again.