On June 30, Anthropic shipped Claude Sonnet 5, a model the company described as delivering “near-Opus performance at a lower price.” The announcement landed with a collective shrug. The AI internet was busy elsewhere — refreshing Polymarket, dissecting a leaked “Honeycomb” build that briefly ghosted through the Cursor editor on July 9, and searching, with growing impatience, for a product called Claude Opus 5.
That product does not exist. It has never been announced. Anthropic’s official models page lists Opus 4.8, Sonnet 5, and — sitting above both — something called Fable 5, a “Mythos”-class model that occupies a tier the company has barely explained. The flagship the market is demanding has already been demoted. Nobody seems to have noticed.
The Tier Shuffle Nobody Covered
Anthropic’s product hierarchy has undergone a quiet reorganization over the past six months. Opus, once the undisputed top of the lineup, now occupies an awkward middle ground. Sonnet 5, released four weeks ago, matches or approaches Opus 4.8 on most benchmarks while costing less. Fable 5, meanwhile, sits in a “Mythos” tier above both — available, according to scattered user reports, but gated behind usage credits and largely absent from the company’s public-facing product pages.
This is not the behavior of a company that plans to crown a new Opus king. It is the behavior of a company that has already moved on from the Opus-as-flagship model and is waiting for the market to catch up. The prediction markets, to their credit, have sensed something is off. Contracts for a 2026 Opus 5 release repriced in early July — not toward certainty, but toward December. The market is pushing the date out, not pulling it in. That is a tell.
Anthropic has shipped four Opus point-releases since November 2025 — 4.5, 4.6, 4.7, and 4.8 — each a meaningful improvement, each greeted with diminishing attention. The company is iterating rapidly on a product line it appears to be phasing out of the top slot. That is not a contradiction. It is a transition.
The Cargo Cult of Version Numbers
What makes the Opus 5 frenzy instructive is not what it says about Anthropic’s roadmap. It is what it says about how the AI industry consumes product news. The public has internalized a model of progress that looks like smartphone releases: a new number every year, a clear hierarchy, a flagship that sits unambiguously at the top. When that pattern breaks — when a company starts shipping sideways, renaming tiers, and slotting new models above old flagships without a press tour — the audience keeps searching for the old pattern anyway.
“Half the people in my Slack channels are still asking when Opus 5 drops,” one engineer at a mid-sized AI startup told me over a direct message this week. “I keep pointing them to the Fable 5 docs. They don’t want to read them. They want the number.”
This is not irrational. Version numbers are a coordination mechanism. They tell procurement teams what to budget for, developers what to build against, and investors what to model. When a company like Anthropic muddies that signal — shipping Opus 4.8 in May, Sonnet 5 in June, and Fable 5 somewhere in between, with no clear statement about what “5” even means anymore — the coordination breaks. The market fills the vacuum with rumors, and the rumors acquire their own gravity.
The Real Product Is the Tier Structure
Anthropic’s actual strategy, if you read the tea leaves of its shipping cadence, is not hard to discern. The company is building a multi-tier product lineup where “Opus” is no longer the ceiling. Sonnet handles the volume. Opus serves the power users who need reliability over novelty. Fable — and whatever follows it — is where the frontier actually lives.
This is, in many ways, a more mature product strategy than the annual-flagship model. It is also harder to communicate. Apple can put a new iPhone on stage and the world understands the hierarchy instantly. Anthropic is asking its users to track three model families, each with their own version numbers, plus an unannounced “Mythos” tier that may or may not be generally available. The confusion is not a bug of the rollout; it is a feature of the strategy.
Whether that strategy works depends on whether the market can be trained to stop searching for Opus 5 and start asking what Fable 5 actually does. So far, the search traffic suggests the answer is no. The Hacker News thread that lit up this week has 711 comments and counting — nearly all of them about a product that Anthropic has never confirmed exists.
The irony is that Anthropic has already shipped the thing the market claims to want: a model more capable than Opus 4.8. It just didn’t call it Opus 5. The name, it turns out, matters more than the product. And until Anthropic says otherwise, the market will keep searching for a ghost of its own making.