Claude Opus 5: Features, Pricing & What's New (2026)

Claude Opus 5 is the newest flagship in Anthropic's Opus line, released in July 2026. Anthropic positions it as coming close to the frontier intelligence of Claude Fable 5 across coding, knowledge work and complex reasoning, at about half the price.

Notably the price did not move from the previous generation, so it is a capability jump with no cost increase. Here is the spec at a glance.

Three things stand out.

First, the capability-to-price ratio: Opus 5 performs close to Claude Fable 5 while costing about half as much, and it holds the same pricing as Opus 4.8, so the upgrade is effectively free if you were already on Opus.

Second, Anthropic reports substantial gains in self-verification and deliberate iteration, meaning the model checks its own work and refines rather than committing to a first answer, which matters most on long agentic tasks.

Third, it ships with a fast mode and a control that lets you trade cost against capability per task, so you are not paying frontier prices for routine work.

Anthropic also describes it as its most aligned model to date.

The useful comparison is capability against cost, because that is the axis Opus 5 moves. Fable 5 remains Anthropic's most capable model, but Opus 5 closes most of the gap for half the money, and Opus 4.8 is now superseded at the same price.

For most teams Opus 5 is now the default choice: you give up a little at the very top end and halve the bill. Reach for Fable 5 only when a task genuinely needs the maximum.

Opus 5 suits work where quality matters but frontier pricing is hard to justify: production coding and refactors, long agentic runs that need the model to verify its own output, research and analysis, and knowledge work with a lot of context, helped by the 1M token window.

The self-verification improvements make it a better fit than earlier models for tasks that run unattended, because it is more likely to catch its own mistakes rather than confidently finish something wrong.

The cost-versus-capability control also makes it practical to run at scale, dialing effort down for routine calls and up for the hard ones.

Opus 5 is available in the Claude apps (the default on Claude Max, and the strongest model for Claude Pro) and through Anthropic's API.

If what you actually want is finished output rather than raw model access, FluxGravity, FluxNote's grounded AI agent, runs on current frontier models, and FluxNote turns a prompt into a narrated, captioned video without any API wiring.

Start free with FluxNote and turn a prompt into a finished, captioned video today.