Anthropic dropped two models on June 8, 2026. One is for everyone. The other is for a very small group of people who protect critical infrastructure. They share the same core, but how you can use them is completely different.
The Safeguard System (Read This First)
Fable 5 has a fallback built in. When you ask about cybersecurity, biology, chemistry, or distillation, the model quietly switches to Claude Opus 4.8 instead of answering from full capability. You probably won’t notice it happening. Less than 5% of sessions ever trigger it, so for the other 95%, you’re getting the full model with no restrictions.
Anthropic is upfront that this is temporary architecture. The fallback exists because Fable 5 is genuinely capable of doing damage in those areas. They plan to tighten the filter over time so fewer legitimate questions get caught. Right now, it’s set conservatively.
What Fable 5 Is Actually Good At
Software engineering is where it pulls ahead most clearly. It outperforms Opus 4.8 by more than 10% on some benchmarks, and the gap gets bigger on longer, more complex work. If you’re running multi-step development workflows or debugging across a large codebase, the difference is real.
On research, it can run autonomously for days without you checking in. Complex analysis, synthesizing sources, extracting information from technical documents — it handles all of it without much hand-holding. For vision tasks, it reads charts, diagrams, and schematics at a level that hasn’t been publicly available before from Anthropic.
The one area it doesn’t fully deliver on for general users is cybersecurity. If that’s your work, keep reading.
Mythos 5 — Same Model, Different Rules
Mythos 5 is Fable 5 with the restrictions removed in some areas. It’s only accessible through Project Glasswing, which is Anthropic’s program for cyber defenders and critical infrastructure providers. If you’re not already in that group, it’s not available to you yet, though Anthropic says they’ll expand access gradually.
The purpose is specific: let authorized defenders use full cybersecurity capability without the fallback kicking in. Vulnerability analysis, threat detection, risk assessment at the deepest level. Those users get Mythos-class output without Opus 4.8 stepping in. General users don’t.
Pricing and Access
The API costs $10 per million input tokens and $50 per million output tokens. Both models are priced the same. Fable 5 is included in paid Claude plans at no extra charge through June 22, 2026. After that, standard plan terms apply.
It’s available on the API, through Claude Code, in paid subscriptions, and on Google Cloud’s Agent Platform. Mythos 5 sits behind the Glasswing program and isn’t publicly accessible yet.
How This Compares to What Came Before
Fable 5 is the most capable model Anthropic has ever made available to the public. The previous ceiling was Opus 4.8, which is now the fallback model when Fable 5 backs off. A private version of the Mythos-class model circulated among selected partners about a month before June 8, which apparently caused some noise in certain technical circles. The public release is the first time this capability tier has been generally available.
Who Should Pay Attention
Developers get a genuine upgrade on complex coding work. Researchers get a model that can carry long autonomous tasks. Businesses using the API or cloud services get production-ready performance. Cybersecurity professionals need Glasswing access to unlock what this model can actually do in their domain.
If you’re on a paid Claude plan, try it now. The free promotional period runs out June 22.
Quick Verdict
Fable 5 is worth switching to immediately. The safeguard fallback affects less than 5% of usage and won’t touch most workflows. For cybersecurity work at full depth, wait for Glasswing access or apply for it. The pricing is unchanged from Opus 4.8, so there’s no reason to stay on the older model.