Anthropic Launches Claude Opus 5: Near-Frontier Intelligence at Half the Price

Anthropic launched Claude Opus 5 on July 24, 2026, and the pitch is unusually blunt about economics: near-frontier intelligence — close to the company's top Fable 5 model — at roughly half the price. After a stretch where the AI story was defined by ever-bigger, ever-pricier flagships, Opus 5 reframes the question from "how smart" to "how much intelligence per dollar."

What launched

Per Anthropic and day-one coverage, Opus 5:

• Comes close to Fable 5's capability while costing about half as much, and is priced the same as its predecessor, Opus 4.8.
• Ships a cost/capability toggle — an effort setting (low → max) that lets users spend more compute on hard problems or run lighter and cheaper on routine work.
• Is the new default on Claude Max, the strongest option on Claude Pro, and the default model in Claude Code.
• Posts state-of-the-art results on coding and knowledge-work evaluations, by Anthropic's measure — though early reports note it isn't top of the table everywhere (it reportedly trails on some specialized areas such as cybersecurity).

We flagged this model while it was still speculation in the Opus 5 / Fable 5 subscription rumor; it's now official.

Why the "toggle" is the interesting part

The effort dial is more than a convenience. It turns a single model into a spectrum: cheap-and-fast for the 80% of routine prompts, expensive-and-deep for the 20% that need it. That's a direct answer to the biggest complaint about frontier models — that you pay flagship prices for work a lighter model could do. Handing users the trade-off explicitly is a smarter cost story than another flat, all-or-nothing tier, and it fits the industry's broader pivot toward intelligence-per-dollar rather than raw capability at any cost — the same pressure reshaping the AI chip race and the search for cheaper inference.

The competitive picture

Half-price near-frontier capability lands squarely in a market where the floor is rising fast. Chinese labs are pushing multi-trillion-parameter, often open-weight models that compete on abundance and cost, while Anthropic keeps its own top tier closed (see Anthropic redeploying Fable 5). Opus 5 is the counter-move: keep the very best behind a wall, but make near-best dramatically cheaper — pulling more usage onto Anthropic's stack without discounting the flagship.

The honest caveats

• The SOTA claims are Anthropic's own. Independent benchmarks will decide how "close to Fable 5" really holds across tasks.
• Not a universal win. "Matches many capabilities" leaves gaps; treat it as a strong generalist, not a leader in every niche.
• Cheaper per token, but the effort dial cuts both ways — Max-effort runs cost more, so the savings depend on how it's used.

Bottom line

Opus 5 is less a raw capability leap than a pricing and packaging move: bring frontier-adjacent quality to a much lower price point, and let users dial the cost/capability trade-off themselves. In a market increasingly won on economics rather than benchmarks alone, that may matter more than another point on a leaderboard.

Reporting: Anthropic, Bloomberg, CNBC, Fortune. Benchmark and pricing claims are as stated by Anthropic at launch; independent evaluations were pending.