Fable 5 vs Mythos 5: Same Model, Different Safeguards
| Spec | Claude Fable 5 | Claude Mythos 5 |
|---|---|---|
| Tier | Mythos-class (above Opus) | Mythos-class (above Opus) |
| Availability (general) | Suspended (export-control directive) ↗ | Suspended — Glasswing only ↗ |
| Released | June 9, 2026 ↗ | June 9, 2026 ↗ |
| Context window | 1M tokens | 1M tokens |
| Max output | Up to 128k tokens | Up to 128k tokens |
| Input price (per M tokens) | $10 / M (after intro window) | $10 / M (reference) |
| Output price (per M tokens) | $50 / M (after intro window) | $50 / M (reference) |
| Data retention | 30 days | 30 days |
| Zero-data-retention (ZDR) | No ZDR | No ZDR |
| Covered Model | Yes — Covered Model | Yes — Covered Model |
| Safeguards | Classifiers for cyber, bio/chem, distillation; reroute to Opus 4.8 | Classifiers lifted (vs Fable 5) |
| How to access | Claude API, AWS Bedrock, Google Cloud, Microsoft Foundry (when not suspended) | Project Glasswing only |
One model, two trust postures
Fable 5 and Mythos 5 are not competing models. They are two presentations of the same June 9, 2026 Mythos-class release, with the same 1M-token context, up to 128k output tokens, and the same $10/$50 per-million pricing after the intro period. If you compared raw weights, they would be indistinguishable.
What differs is the safety layer wrapped around the model. Fable 5 ships with classifiers covering cybersecurity, bio/chem, and model distillation. When a request trips one, that session is rerouted to Opus 4.8 — this affects under 5% of sessions, and you are told when it happens. Mythos 5 removes that reroute entirely and runs the unmediated model, which is why it is gated behind Project Glasswing vetting.
When you'd pick which
You usually don't pick. General users get Fable 5; vetted Glasswing organizations get Mythos 5. Choose Mythos 5 only if classifier fallback is genuinely disrupting legitimate research and you qualify for Glasswing. Otherwise Fable 5 is the same capability with a safety net.
Caveat
Both editions currently have access suspended under a June 12, 2026 export-control directive, so neither is broadly reachable today. Any capability "benchmarks" you see circulating are unofficial until Anthropic publishes a system card.
Frequently asked questions
Is Mythos 5 a smarter model than Fable 5?
No. They are the same model. Mythos 5 may complete a few prompts that Fable 5 reroutes, but that is a safeguards difference, not a capability upgrade.
Will I notice the classifiers in Fable 5?
Most users won't. Fewer than 5% of sessions are rerouted to Opus 4.8, and Fable 5 tells you when a reroute happens rather than failing silently.
Can I just choose Mythos 5 to avoid reroutes?
Only if you are an approved Project Glasswing participant. Mythos 5 is not offered on standard plans.
Sources & further reading
- Anthropic — Fable & Mythos access noticeOfficialanthropic.com
- anthropicmythos.aiFlagshipanthropicmythos.ai
Facts on this page link to their source. Quotes are kept under 15 words and attributed; figures labelled unofficial are third-party until Anthropic publishes system-card numbers.