Content is user-generated and unverified.
Learn how to customize
Anthropic · Released July 24, 2026
Claude Opus 5 ships with a dial on the bill.
Anthropic’s new flagship costs exactly what Opus 4.8 cost. What is new is that you decide how hard it works on any given task.
The effort setting
Low effort for sorting, tagging, and first drafts. High or max for the board memo, the grant narrative, the analysis a funder will read. More effort means more tokens spent, so the dial is a budget control as much as a quality control. Bar heights are illustrative.
Fewer tokens on average to match Opus 4.8 at maximum reasoning, according to legal AI firm Harvey.
Opus 5 is now the default model on Claude Max and the strongest option on Claude Pro.
Anthropic calls it its most aligned Opus model to date, with the lowest measured rates of deceptive behavior.
$25 / million output tokens
Unchanged from Opus 4.8
ExecSearches.com · by JAY
Claude Opus 5 Is Here, and the Price Did Not Move
Anthropic released Claude Opus 5 on July 24, 2026. If you have been running your grant drafts, donor letters, and job descriptions through Opus 4.8, the short version is this: you now get a meaningfully better model at exactly the same price, and you get a new control over what each task costs you.
What actually changed
Three things matter for a small mission-driven team.
- An effort dial. Opus 5 ships with an adjustable effort setting that runs from low through medium, high, xhigh, and max. Low effort is fine for tagging a list of prospects or knocking out a first draft. High or max is what you want on the analysis that goes in front of a board or a program officer. Higher effort spends more tokens, so this is a budget lever, not just a quality lever.
- More work per dollar. Anthropic’s published charts are framed around score versus cost per task rather than peak score alone. The legal AI company Harvey reported that Opus 5 matched Opus 4.8’s maximum-reasoning performance while generating about 26 percent fewer tokens on average.
- A steadier model. Anthropic describes Opus 5 as its most aligned Opus model to date, with the lowest measured rates of deceptive or reckless behavior of any recent Claude. For anyone handing an AI a funder relationship or a candidate writeup, that is the part worth reading twice.
What it is not
Opus 5 is not the top of Anthropic’s lineup. The company still points to Claude Fable 5 for the most advanced projects, including work a model might run autonomously for days, and Opus 5 deliberately trails Anthropic’s security-focused Mythos 5 on high-risk technical work. Think of Opus 5 as the everyday flagship: the model you reach for by default, not the absolute ceiling.
What this means if you are already using Claude at work
Nothing breaks. If you are on Pro or Max, Opus 5 is already in your model picker. The practical move this week is to take one workflow you already trust Claude with, a funder research brief or a candidate summary, run it again on Opus 5, and compare. Then start paying attention to the effort setting instead of leaving it wherever it lands.
We will be walking through that comparison in detail in the next few posts, including a side-by-side against Opus 4.8 and a practical guide to choosing an effort level for common nonprofit and recruiting tasks.
