Anthropic’s Literary Model Names—from Aphorism to Cinematic Universe—Expose Product Issues

Yesterday a satirical post on Hacker News by Sam Wilkinson humorously "predicted" Anthropic’s naming roadmap, progressing from Aphorism to Cinematic Universe (Director’s Cut).

This table is more serious than Anthropic’s official site

Wilkinson paired the "Claude Fable release" with the tagline: "Anthropic is evolving from a single poem to an enterprise‑level narrative unit." The table he created includes:

Aphorism : a one‑sentence statement that is always spot‑on

Haiku : a short poem, a small bill

Marginalia : unsolicited annotations on your code

Abstract : a summary of reasoning the model never performed

Sonnet : medium‑length poem, medium‑length bill

Diatribe : an angry Sonnet

Opus : a long poem, a long bill

Treatise : Opus, but citations left for the reader

Mythos : a horror‑style Opus

Fable : Mythos, until the question really matters

Fable (xhigh) : bankruptcy‑fast track

Saga (Unabridged) : answers questions you didn’t ask

Lore : requires a wiki to interpret

Cinematic Universe : multiple Sagas with an extra layer of Lore scheduling

Terms of Service : no responsibility for answers or consequences

The full table also contains entries such as Overwhelmingly Large Narrative Unit and Omnibus, which illustrate the naming spiraling out of control.

The humor lies in realizing that the author is mocking a product decision: when a company’s model names become increasingly literary, its product matrix is often already unmanageable.

Anthropic’s naming started with intention

In 2024 the Claude 3 series debuted with three sizes named Haiku, Sonnet, and Opus. Haiku evokes a three‑line, seventeen‑syllable poem; Sonnet references a fixed fourteen‑line form; Opus denotes a composer’s major work. This metaphorical progression from short to long feels more sophisticated than the numeric ladders used by GPT‑3.5, 4, 4o, 4.1, 4.5, 5, 5.1.

At that time no one complained about the naming.

The problem emerged over the next year and a half. Versions like 3.5 Sonnet (New), 3.7 Sonnet, Claude 4 Opus, and Claude Sonnet 4.5 began overlapping, with version numbers, tiers, and sub‑versions covering each other. Even Anthropic’s own website had to redraw a comparison chart.

Looking back at Wilkinson’s meme table, the line "Fable, until the question matters" is not merely a joke—it points to the mid‑tier model that seems impressive but cannot safely meet SLA commitments.

Naming loss of control is a product problem, not a copy issue

OpenAI’s Sam Altman has admitted that GPT’s naming is messy, and Google’s Gemini Pro/Ultra/Flash/Nano have undergone similar revisions. Anthropic’s case, however, warrants special attention.

Literary naming incurs two hidden costs:

It decouples from capability dimensions . Users can infer that Haiku is short and Opus is long, but they cannot tell whether Sonnet offers the best price‑performance or serves as a mid‑range fallback without consulting the website, which defeats the purpose of a name.

It forces every new version to find a new term . When Claude 5 Sonnet exists, what do we call Claude 6 Sonnet? Continue stacking terms or switch to a new literary form? Wilkinson’s “Saga / Lore / Cinematic Universe” escalation illustrates the absurdity of this path.

Model naming is essentially a contract of promise and delivery. OpenAI uses numbers to promise continuity; Anthropic uses poetic forms to promise hierarchy. The former breaks when numbers become endless; the latter breaks when the poetic lexicon runs out.

To solve the issue, Anthropic could revert to an engineering‑style label such as "name + parameter count + version" (as Mistral or DeepSeek do) or simplify the lineup like Apple’s Pro/Air/Standard tiers.

The joke hides a real question Anthropic must answer

You can read Wilkinson’s table as pure humor, or as an issue ticket stating that users can no longer discern which model fits which use case.

Claude’s current product strength is evident—Sonnet 4.5 leads its class on coding benchmarks—but when selecting between claude-sonnet-4-5-20250929, claude-3-5-sonnet-20241022, and claude-opus-4-5-20251101, the anxiety mirrors the “wrong version installed” panic of an npm install.

Omnibus : fine‑tuning continues until morale improves.

Translated, this means that if naming continues down this path, the first to collapse won’t be Claude—it will be the user.

In short, Anthropic’s most urgent need isn’t the next Opus; it’s a clear, human‑readable product comparison table.

When you last upgraded your Claude model, was it based on benchmark results, price, or simply inertia? Share your thoughts in the comments.