senn-techAI Radar: Claude Fable 5 and Mythos 5
Anthropic released its strongest model generation to date on June 9, 2026 — and took an unusual approach: instead of a single “Claude 5", the new Mythos class appears as two products.
One Model, Two Safety Profiles
- Claude Fable 5 — generally available and, according to Anthropic, the most powerful openly accessible model they have ever released. Safety classifiers automatically reroute sensitive requests (cyber, bio/chemistry, model extraction) to Claude Opus 4.8 — according to Anthropic without noticeable impact: in over 95 % of sessions, no single fallback is triggered.
- Claude Mythos 5 — the same model with relaxed safeguards, initially only for vetted partners (cybersecurity program “Project Glasswing", later selected researchers).
Where It Excels
Fable 5 is, according to Anthropic, “state of the art" across nearly all benchmarks — particularly strong in software engineering, knowledge work, and scientific research. It thus takes the lead over the existing Claude-4 family (Opus, Sonnet, Haiku).
Pricing
$10 per million input tokens, $50 per million output tokens — according to Anthropic less than half the Mythos Preview. Fable 5 is included at no extra cost in the subscription plans (Pro, Max, Team, Enterprise) until June 22, after which it is available via credits or the API.
Our Take
What’s interesting is less the raw performance than the safety design as a product feature — built-in guardrails instead of after-the-fact filters. The caveat remains the same as with any cloud API: data leaves the premises. For confidential or personal-data workloads, we combine cloud models with self-hosted on-prem models in the EU — deciding which model for which use case, we plan this together with our customers.
Further Reading
What is the difference between Claude Fable 5 and Mythos 5?+
Both are based on the same model but differ in their safety profile. Fable 5 is generally available; safety classifiers automatically reroute sensitive requests to Claude Opus 4.8 — in over 95 % of sessions no fallback triggers. Mythos 5 has relaxed safeguards and is initially only for vetted partners such as the Project Glasswing cybersecurity program.
What does Claude Fable 5 cost, and from when is it billed?+
The API costs $10 per million input tokens and $50 per million output tokens — according to Anthropic less than half the Mythos Preview. Until June 22, Fable 5 is included at no extra cost in the Pro, Max, Team, and Enterprise subscription plans; afterwards it runs via credits or the API.
Can we run Claude Fable 5 on-premises?+
No. Like the entire Claude family, Fable 5 is proprietary and API-only — no open weights, no self-hosting. Data leaves the premises. For confidential or personal-data workloads we therefore combine cloud models with our own on-prem models in the EU and decide case by case which model is used for what.