Decoding Anthropic's Model Names: Fable, Mythos, and What the Naming Shift Signals

A satirical table called "Anthropic's Model Naming, Extrapolated" hit 204 points on Hacker News today. The joke is that Anthropic's naming convention - already the most poetic in the industry - now stretches far enough to extrapolate a full literary canon: Saga, Canon, Lore, Cinematic Universe, all the way to "Zack Snyder's Saga (terminal turns black and white, becomes harder to follow)." It landed because the underlying observation is real: with Claude Fable 5, Anthropic crossed a line it had not crossed before. The old tier hierarchy - Haiku, Sonnet, Opus, each with a version number - was easy to reason about. Now there is a named class sitting above Opus, and the naming logic has shifted in a way that matters for anyone writing code against the API.

Last updated: June 10, 2026

The Old Ladder and Why It Worked

For most of Claude's public history, the naming system was simple enough to fit in a README comment. Three tiers, ordered by capability and cost:

• Haiku - fast, cheap, best for high-volume tasks where speed and cost dominate
• Sonnet - the middle ground, strong reasoning at a reasonable price point
• Opus - the flagship, highest capability, highest price

Version numbers tracked iteration within each tier. Claude 3 Opus, Claude 3.5 Sonnet, Claude 4 Opus. You could look at a model ID and immediately understand two things: the capability class and how recently it was updated. The system was not perfect - the version numbers sometimes jumped in confusing ways, and "Sonnet" did not always mean the same thing across generations - but the mental model held.

The community noted this favorably. As one HN commenter put it today, comparing it to OpenAI's lineup of o3, 4o, 4o-mini, o4-mini, gpt-4.1, gpt-4.1-mini, and gpt-4.5 (Research Preview): "I miss the days when the dropdown asked me to choose between all of those." Anthropic's names were a small mercy.

What Changed with Mythos and Fable

With Claude Fable 5, Anthropic introduced something structurally new: a named capability class called Mythos, positioned above Opus. Fable is the first model in that class. For the practical decision between the two top tiers, see Fable 5 vs Opus 4.8: when to use which.

This breaks the prior pattern in two ways.

First, the tier name is now a class name, not a poetic register. Haiku, Sonnet, and Opus are all forms of writing - they describe something about the model's style of output, loosely speaking, or at least they exist in the same literary register. Mythos is a category. It describes what the tier is for, not what it sounds like.

Second, the top tier now has a model name distinct from the class name. Fable 5 lives inside Mythos class the way a product lives inside a product line. The model ID claude-fable-5 does not contain the word "Mythos." Mythos is the capability bucket; Fable is the named flagship inside it.

What does that separation buy Anthropic operationally? Possibly a great deal. It lets them release a second Mythos model - say, one optimized for agentic tasks or with different safety characteristics - without it being "Fable 5.1." The class absorbs the positioning; the name carries the personality.

What "Mythos Class" Means in Practice

Mythos is not just a marketing label. Based on the Fable 5 release details, the class comes with distinct operational characteristics:

Pricing is in a different band. At $10 per million input tokens and $50 per million output tokens, Fable 5 is roughly 2x the cost of Opus 4.8 on input and 2x on output. That gap is significant enough that you should not default to it without a reason.

Context window extends to 1M tokens, which puts it in a different class of problem. Tasks that require holding an entire codebase, a full legal document set, or a long conversation history in context are now viable in a single call rather than requiring chunking logic.

Safety and evaluation requirements at the Mythos tier are understood to be more rigorous than at Opus. Anthropic's internal ASL (AI Safety Level) framework gates model releases on evaluation benchmarks that grow stricter as capability increases. A model above Opus is, by that framework, a model that required new gates to clear. What those gates look like for external API users in terms of rate limits, use-case restrictions, or enterprise agreements is something worth watching as the tier matures.

Model Tier Reference

The practical decision tree for most teams is: start at Sonnet 4.6, move to Opus 4.8 when Sonnet fails on your evals, and reach for Fable 5 only when Opus fails or when the task genuinely requires 1M-token context. The price step from Sonnet to Fable 5 is roughly 17x on output tokens. That is not a rounding error in a production budget.

The Naming Signals Something About Roadmap

Names are roadmap signals. When Anthropic introduced a class above Opus, it was implicitly committing to a world where Opus is not the ceiling. That has downstream implications worth thinking through.

Does Opus get discontinued as the Mythos tier matures? Probably not soon - Opus 4.6 and 4.7 still exist alongside 4.8, and there is clearly a strategy of keeping older checkpoints available for teams that have tuned their prompts against specific versions. But the pressure on Opus as a premium positioning is real. If Fable 5 becomes the reference for "best available," Opus 4.x becomes the sensible default for teams that want near-flagship quality without flagship pricing.

The version numbers on Opus are also worth watching. Opus 4.8, 4.7, 4.6 are a fast iteration cadence. Are those training improvements, safety re-evaluations, or capability regressions in specific areas? Anthropic's model cards have historically been informative but not exhaustive. Knowing which Opus you are on matters when you are comparing eval results across runs.

Open Questions the Community Is Already Asking

The HN thread today spent most of its energy on the joke - suggestions for future names included Saga, Canon, Lore, Cinematic Universe, and (my favorite) "Zack Snyder's Saga: same answer, terminal turns black and white." But a few comments surfaced real questions that do not yet have answers.

One commenter noted that the original Haiku and Sonnet names may have come from a nearby coffee shop (Postscript Coffee, which sells beans named Haiku and Sonnet). That would mean the first two tier names were borrowed, not invented - which makes the literary escalation to Mythos feel more like a deliberate rebrand than a natural extension.

Another thread discussed whether Anthropic is "nerfing" model behavior intentionally - a real concern, separate from naming, but worth flagging because it shapes how people interpret the tier system. If a model in a higher tier behaves more conservatively on certain task types due to safety gating, the capability ordering is not monotonic for all use cases.

The questions that matter most for developers building on the API right now:

• Will there be a Fable 5.1, or does the named-model convention mean major capability jumps only?
• Will a second Mythos model appear with a different name - and if so, how do you compare two named models in the same class?
• What happens to Opus version numbering as Fable 5 matures? Does Opus 5.0 exist, or does the Opus line stay in the 4.x range permanently?
• Does the Mythos class come with different API terms - enterprise agreements, usage policies, or rate limit structures - that do not apply to Opus and below?

None of those have definitive public answers today. They are worth tracking in release notes and API changelogs over the next few months.

How to Make Decisions Now

The practical upshot for most development teams is straightforward. The naming change does not make model selection harder - it gives you more vocabulary. Mythos class = maximum capability, 1M context, highest price. Opus = strong capability, standard context, sensible premium price. Sonnet = workhorse. Haiku = cheap and fast.

What has changed is that the tier above Opus now has a name rather than just a version number, and that name signals Anthropic's intention to invest in this class as a distinct product line rather than just incrementing Opus. Whether that investment manifests as more frequent named-model releases, enterprise-specific features, or something else is an open question.

For now: if you are evaluating whether to move production workloads to Fable 5, the useful question is not "is it better than Opus 4.8?" - the answer is yes on most benchmarks. The useful question is whether your specific task benefits from the 1M context window or justifies the 2x price premium. If neither is true, Opus 4.8 or Sonnet 4.6 is almost certainly the right call.

FAQ

What is Claude Mythos?

Mythos is Anthropic's top capability class, sitting above Opus in the model hierarchy. Claude Fable 5 is the first model in the Mythos class. The class is characterized by the highest capability tier, a 1M-token context window, and premium pricing ($10/$50 per million tokens input/output). Think of Mythos as the product line name and Fable as the specific model within it.

Is Fable 5 better than Opus 4.8?

On most general reasoning and long-context tasks, yes. Fable 5 also supports a 1M-token context window versus Opus 4.8's 200K. However, "better" depends on your task. Fable 5 costs roughly 2x as much per token as Opus 4.8. If your workload fits within 200K tokens and your evals show Opus 4.8 meeting your quality bar, Fable 5 adds cost without adding value. Test on your actual tasks before switching production traffic.

What model IDs should I use in my API calls?

Use claude-fable-5 for Mythos-class tasks, claude-opus-4-8 for the current Opus flagship, claude-sonnet-4-6 for most production workloads, and claude-haiku-4-5 for high-volume low-cost tasks. Avoid pinning to generic aliases like claude-opus-latest in production - pin to the specific version ID so you control when upgrades happen.

Why did Anthropic introduce a named tier above Opus?

The most likely reason is that a named tier gives Anthropic product flexibility that version numbers do not. With a named class, Anthropic can introduce multiple models within Mythos (each with different characteristics) without those models being forced into an Opus version sequence. It also signals to enterprise customers that the highest tier is a deliberate product investment, not just an incremented model. The community has noted that the literary naming - Haiku, Sonnet, Opus, now Fable under Mythos - is more memorable and differentiating than the version-number soup that competitors use.

Will there be more named models like Fable?

Possibly. The structure of the Mythos class - a tier name plus a model name inside it - is set up to accommodate multiple named models. Whether Anthropic releases a "Claude Saga" or another Mythos-class model with a different profile (say, one tuned for agentic tasks versus long-context synthesis) is an open question. The HN thread today was full of half-joking predictions: Saga, Canon, Lore, Chronicle. Some of those are plausible. Watching Anthropic's model release cadence and API changelogs is the best way to track this.

---

Sources

• Sam Wilkinson, "Anthropic's Model Naming, Extrapolated" (June 9, 2026) - https://samwilkinson.io/posts/2026-06-09-anthropics-model-naming-extrapolated
• Hacker News discussion, story #48480852 (June 10, 2026) - https://news.ycombinator.com/item?id=48480852
• Anthropic API documentation, model overview - https://docs.anthropic.com/en/docs/about-claude/models/overview
• Claude Fable 5 announcement - https://www.anthropic.com/news/claude-fable-5