Claude Fable 5 & Mythos 5: Anthropic's Mythos-Class Tier, Explained
Anthropic's Claude Fable 5 and Mythos 5 explained: the new Mythos-class tier above Opus, why the same model ships two ways, and what it means for Claude Code.
Anthropic has introduced Claude Fable 5, the first model in the new Claude 5 family — and with it, something the Claude lineup has never had before: a tier that sits above Opus. Alongside Fable 5 comes Claude Mythos 5, which is not a second model at all. It is the same underlying model, shipped under a different access policy.
That last sentence is the easiest part to get wrong, so let’s unpack the announcement properly: what the Mythos-class tier actually is, why one model ships under two names, and what changes for you as a Claude Code user.
What Was Announced
The core facts, stripped of hype:
- Claude Fable 5 (model ID:
claude-fable-5) is the first model in Anthropic’s new Claude 5 family. - It belongs to a new Mythos-class tier that sits above Claude Opus in capability.
- Fable 5 and Mythos 5 share the same underlying model. The split is about configuration and access:
- Fable 5 is Anthropic’s most intelligent generally available model. It includes additional safety measures for dual-use capabilities.
- Mythos 5 is the same model without those measures, available only to approved organizations.
The full details are in Anthropic’s official announcement. Everything below is analysis of what it means in practice.
A New Tier Above Opus
For years, “Opus-class” was shorthand for “the most capable Claude you can get.” The Claude 5 family breaks that assumption. Mythos-class is a tier name — the same way “Opus-class” is a tier name — and it now occupies the top of the ladder:
┌──────────────────────────────────────────────────────────┐
│ Mythos-class (NEW) Fable 5 (GA) / Mythos 5 (gated) │ ← above Opus
├──────────────────────────────────────────────────────────┤
│ Opus-class Opus 4.8 │
├──────────────────────────────────────────────────────────┤
│ Sonnet-class Sonnet 5 │
├──────────────────────────────────────────────────────────┤
│ Haiku-class Haiku 4.5 │
└──────────────────────────────────────────────────────────┘The naming is worth internalizing because it will shape how the family grows: Mythos-class describes where the tier sits; Fable 5 is the generally available model within that tier. When you see “Mythos” in Anthropic’s materials, think tier and access track, not “a bigger secret model.”
Same Model, Two Doors
Here is the comparison that matters:
| Claude Fable 5 | Claude Mythos 5 | |
|---|---|---|
| Underlying model | Same model | Same model |
| Availability | Generally available | Approved organizations only |
| Safety configuration | Additional safety measures for dual-use capabilities | Without those measures |
| Position | Anthropic’s most intelligent GA model | Same tier, gated access |
Two doors into one room. The difference between Fable 5 and Mythos 5 is not parameters, training data, or intelligence — it is what is wrapped around the model and who is allowed through which door.
This is why “is Mythos 5 smarter?” is a malformed question. It cannot be smarter; it is the same model. What differs is that Fable 5 carries extra safety measures targeting dual-use capabilities — capabilities that are legitimate in some hands and dangerous in others — while Mythos 5 ships without those measures to organizations Anthropic has explicitly approved.
Why Split a Tier Instead of Shipping Two Models?
The interesting move here isn’t the capability jump — every frontier release claims one of those. It’s that Anthropic decoupled two questions that used to be answered together:
- How capable is the model? (the tier: Mythos-class, above Opus)
- Who gets which configuration of it? (the access policy: GA with safeguards, or approved-only without)
In previous generations, those were fused: if a capability was too sensitive to ship broadly, the option was to hold the model back or restrict it for everyone. The Fable/Mythos split creates a third option. General users get the frontier model with dual-use safeguards; vetted organizations that need the unrestricted configuration can apply for it. Capability goes out broadly; the sharpest edges stay gated.
For most developers, the practical consequence is simple: Fable 5 is the top of the lineup as far as you’re concerned. The Mythos track exists, but unless your organization goes through Anthropic’s approval process, it isn’t part of your day-to-day tooling — and you’re not missing a “better brain,” because there isn’t one behind that door.
What This Changes for Claude Code Users
A new model at the top of /model
Fable 5 is available in Claude Code as claude-fable-5. Switching is the same in-session flow you already know:
/model # pick the model (e.g. Fable 5, Opus 4.8, Sonnet 5, Haiku 4.5)
/effort # pick reasoning effort for the session
/fast # Opus fast mode: same Opus model, faster output (Opus 4.8/4.7)Both /model and /effort choices can be saved as defaults for new sessions. Note what /fast is not: it does not silently downgrade you to a smaller model — it runs Opus with faster output.
Where Fable 5 earns its keep
Anthropic’s guidance for building AI applications is to default to the latest and most capable Claude models. Within a Claude Code session, the qualitative positioning looks like this:
- Fable 5 — the top generally available tier. Reach for it on the hardest judgment work: architecture decisions, adversarial verification, complex multi-step autonomy, and final synthesis. It is also the model that ships with ultracode-style orchestration guidance, so it pairs naturally with workflow orchestration.
- Opus 4.8 (
claude-opus-4-8) — the frontier Opus-class workhorse, with fast mode for interactive sessions where latency matters. - Sonnet 5 (
claude-sonnet-5) — the balanced default for everyday coding, successor to the Sonnet 4.x line that was long the Claude Code default. - Haiku 4.5 (
claude-haiku-4-5-20251001) — the fast/cheap tier for exploration subagents and mechanical fan-out stages.
Mixing tiers is the real skill
The tier split reinforces a pattern that was already emerging: model choice is moving from per-session to per-task. Claude Code gives you three places to override the model for a specific agent:
.claude/agents/*.mdfrontmatter — pin a model per agent type- The Agent tool’s
modelparam —sonnet | opus | haiku | fable - Workflow
agent()opts —modelandeffortper spawned subagent
The productive pattern: cheap models with low effort for mechanical stages (reading, searching, formatting), the top tier with high effort for verify/judge/synthesis stages. And the sane default is to omit the override entirely and inherit the session model unless you’re confident a different tier fits. We walk through this in depth in the 2026 model selection guide.
The Mid-2026 Lineup at a Glance
| Model | Model ID | Tier | Reach for it when |
|---|---|---|---|
| Fable 5 | claude-fable-5 | Mythos-class (GA) | Architecture calls, adversarial verification, long autonomous runs, final synthesis |
| Opus 4.8 | claude-opus-4-8 | Opus-class | Frontier workhorse; /fast when latency matters |
| Sonnet 5 | claude-sonnet-5 | Sonnet-class | Everyday coding — the balanced default |
| Haiku 4.5 | claude-haiku-4-5-20251001 | Haiku-class | Exploration subagents, high-volume fan-out |
FAQ
Is Mythos 5 a bigger or smarter model than Fable 5?
No. Fable 5 and Mythos 5 share the same underlying model. The difference is configuration and availability: Fable 5 includes additional safety measures for dual-use capabilities and is generally available; Mythos 5 omits those measures and is gated. “Mythos = Fable Pro” is the wrong mental model.
Who gets access to Mythos 5?
Approved organizations only. Everyone else — which means nearly everyone — uses Fable 5, and loses nothing in raw model capability by doing so.
Which one do I use in Claude Code?
Fable 5. It’s the generally available model (claude-fable-5); select it with /model and save it as your default if you want it across sessions.
Does Fable 5 make Opus 4.8, Sonnet 5, and Haiku 4.5 obsolete?
No. This is a portfolio, not a replacement cycle. Opus 4.8 remains the frontier workhorse (with fast mode), Sonnet 5 the everyday default, Haiku 4.5 the fan-out engine. The skill is matching the tier to the task — and mixing tiers within one session.
What’s Fable 5’s knowledge cutoff?
January 2026.
What about pricing, benchmarks, and context window sizes?
We’re deliberately not quoting numbers here. Anthropic’s announcement post is the authoritative source for specs; this article sticks to positioning and workflow implications that are stable regardless of spec-sheet details.
The Bottom Line
Three things to take away from the Fable 5 / Mythos 5 announcement:
- The ladder grew a rung. Mythos-class is a new tier above Opus, and Fable 5 is its generally available model — Anthropic’s most intelligent model you can actually use today.
- Two names, one model. Fable 5 and Mythos 5 differ in safety configuration and access, not intelligence. Don’t chase the gated one.
- Model strategy beats model loyalty. With four active tiers and three per-task override points in Claude Code, the win isn’t “always use the biggest model” — it’s putting Fable 5’s judgment where it counts and cheaper tiers everywhere else.
Want the hands-on version? Work through Session 25: The Claude 5 Family & Model Strategy, or start with our practical model selection guide for 2026.