Anthropic Model Naming Conventions: A Satirical Extrapolation

Anthropic's Naming Strategy: From Poetry to Narrative

Anthropic utilizes a literary-themed hierarchy for its model tiers, currently centered around the triplet of Haiku (small/fast), Sonnet (medium/balanced), and Opus (large/powerful). This naming convention distinguishes the company's branding from the more alphanumeric approach used by competitors like OpenAI (e.g., o1, GPT-4o) or Google (e.g., Gemini 1.5 Pro).

Satirical Projections for Future Model Tiers

Based on the current trajectory of literary and narrative naming, future model expansions could logically extend into larger, more complex formats. The following extrapolated names and their satirical behaviors reflect the perceived evolution of LLM capabilities and costs:

Advanced Narrative Tiers

  • Mythos: An upgraded Opus, described as "scary" in its capability.
  • Fable: A version of Mythos that maintains performance until the complexity of the question increases.
  • Saga: An extension of Fable characterized by "extra meandering" responses.
  • Canon: A model that refuses to contradict its own prior answers to maintain internal consistency.
  • Lore: A model where interpretation requires an external wiki to understand the output.

High-Scale and Experimental Tiers

  • Cinematic Universe: A system utilizing multiple Sagas with a "Canon dispatch layer" to manage consistency across outputs.
  • Overwhelmingly Large Narrative Unit: A model so massive it requires a "previously on" segment before it can be used.
  • Zack Snyder's Saga: A specialized version where the terminal turns black and white and the output becomes harder to follow.

Specialized Utility Tiers

  • Aphorism: A model smaller than Haiku, providing one-sentence responses that "always feel right."
  • Marginalia: A utility model providing unprompted commentary on code.
  • Abstract: A model that summarizes reasoning it has not yet performed.
  • Treatise: An Opus-level model where citations are left as an exercise for the reader.

Community Insights and Counter-Proposals

Users and developers have synthesized further extensions to this naming schema, suggesting that the evolution of AI naming reflects the broader industry's "semantic inflation."

Proposed Literary Additions

Community members suggested several other literary-themed names to fit the existing ecosystem:

  • Free Verse: A proposed name for an open-weight distillation of Anthropic's models.
  • Epic, Tale, Chronicle, and Legend: General suggestions for larger-scale narrative models.
  • Tractatus: A model highly aware of its limitations that numbers its statements and is plagued by a context window bug.
  • Fan Fiction: A term used by some to describe open-weight distillations or riffs on existing models.

Critical Perspectives on Model Performance

While the naming is praised for its aesthetic appeal, some users argue that the branding masks underlying performance issues. One user noted that "Anthropic intentionally degrading responses shows you they're also focusing on the wrong things," suggesting a focus on "nerfing" capabilities rather than maximizing them. Another user expressed concern over the pricing of the Haiku tier, noting that it is not competitive compared to alternatives like DeepSeek v4 Flash.

"The relative personalities of the orgs involved show somewhat in their naming: {o1,4o,3.5,5.5} vs. {Haiku,Sonnet,Opus} vs. {3.1, 3.5-pro, 3.5-flash, 3.5-flash-lite} vs. {3.7, 3.7-plus, 3.7-max}"

The "End-Game" Models

As the industry moves toward Artificial Super Intelligence (ASI), the community suggests names that signify finality or the end of a literary cycle:

  • Requiem: A proposed name for an ASI model.
  • Claude Fin: A play on the French word for "end," suggesting a final, far-ASI monster.
  • Epitaph, Obituary, and Eulogy: Satirical suggestions for models that signify the end of the company's service or the user's relationship with the model.

Sources