Claude Fable 5 & Claude Mythos 5 — System Card
Status — suspended. Access to Claude Fable 5 and Claude Mythos 5 was suspended on June 12, 2026 following a US government export-control directive. Anthropic has stated it is working to restore access. The system card itself remains published — which is why we archive it here.
The document
- 🔗 Canonical landing page: anthropic.com/claude-fable-5-mythos-5-system-card
- 📰 Launch announcement (carries the suspension notice): anthropic.com/news/claude-fable-5-mythos-5
- 📄 PDF (~26 MB, 319 pages): the full system card is kept as a local archive outside the repo (it's a 26 MB binary — too heavy to track in git). Fetch a fresh copy from the canonical landing page above, which redirects to the published PDF.
What the card documents
Two configurations, one model
Fable 5 and Mythos 5 are the same underlying model exposed under two configurations:
- Fable — stricter, built-in safeguards. The default, broadly available configuration.
- Mythos — fewer constraints in some areas, with correspondingly tighter access controls.
Safeguard architecture
- Domain routing. Queries flagged by the cyber and biology safeguards are automatically routed to Opus 4.8 instead of being answered by Fable.
- Transparency. Users are informed whenever this redirection occurs.
- Frequency. Anthropic expects routing to trigger in fewer than 5% of interactions with the model.
- Data retention. Using Fable requires 30-day data retention for safety monitoring.
Alignment results
- Mythos 5's measured level of misaligned behavior was low, and similar to that of Opus 4.8.
- The card includes automated alignment-evaluation metrics demonstrating comparable performance to earlier models, alongside the full suite of safeguard evaluations.
Capabilities
State-of-the-art on nearly all tested benchmarks, with exceptional performance in:
- Software engineering / agentic coding
- Knowledge work
- Scientific research
- Vision and computer use
Why this lives in our docs
Bike4Mind's architecture treats the frontier model as a swappable engine behind a disciplined propose-approve pipeline. When an engine is pulled out from under us — by deprecation, suspension, or regulation, as happened here — the contract we relied on (its safeguards, routing behavior, alignment posture, and capability envelope) still matters for everything we built on top of it. The model card is that contract in writing, so we keep it.