Anthropic Launches Claude Opus 5

Anthropic released Claude Opus 5 on 24 July 2026, the newest version of its top tier Opus model line, and it is available immediately through the Claude API, Claude.ai, Claude Code and Claude Cowork. It is now the default model on Claude Max and the strongest model offered on Claude Pro.

In its announcement, Anthropic describes Opus 5 as "a thoughtful and proactive model that comes close to the frontier intelligence of Claude Fable 5 at half the price." Pricing is set at 5 dollars per million input tokens and 25 dollars per million output tokens, the same rate as its predecessor, Opus 4.8, despite the jump in capability.

Anthropic's own benchmark claims put Opus 5 at state of the art on Frontier-Bench and GDPval-AA, two evaluations covering coding and knowledge work. On ARC-AGI 3, a test built around novel problem solving, the company says Opus 5 scores three times as high as the next best model. On CursorBench 3.2 at maximum effort, it lands within 0.5 percent of Fable 5's peak score at half the cost per task. The company is upfront that it still trails Mythos 5 on cybersecurity exploit tasks.

The pattern across all of this is efficiency rather than a pure intelligence jump: Anthropic's own charts show Opus 5 beating Opus 4.8 by a wide margin at the same price point, not simply scoring higher at a higher cost.

For studios like ours, the practical question is less which model is smartest and more which model stays useful across a long task without drifting off course. Long running agent work, multi step build pipelines and coding assistance that holds context over hours rather than minutes are exactly where Anthropic is positioning Opus 5, and where older models tended to lose the thread. We will be testing it against our own agent workflows over the coming weeks.