Safety, Access, and Cost

How sensitive queries route to other models, what Fable 5 costs, and when to reach for it.

Releasing a Mythos-class model publicly raised an obvious question: how do you make the most capable model safe without making it less capable? Anthropic's answer in Fable 5 is routing, not capping.

Instead of limiting what the model can do, sensitive queries are automatically routed to other systems:

Cybersecurity queries go to Claude Opus 4.8.

Biology and chemistry have protections for dual-use research (research that could be used for good or for harm).

Distillation protections guard against attempts to copy the model's capabilities.

In practice you will rarely notice any of this. The routing triggers in less than 5 percent of sessions. Around 95 percent of sessions run entirely on Fable 5.

Access and cost. Through June 22, 2026, Fable 5 is free for Pro, Max, Team, and Enterprise users. After that, using it requires computing credits. API pricing: $10 per million input tokens, $50 per million output tokens.

When to reach for it? Fable 5's advantage grows with task length and complexity, so:

Use Fable 5 for long, messy, multi-step work: big research projects, large document synthesis, full project builds.

Use a regular Claude model for quick questions, short emails, and everyday tasks. Save Fable 5 for the work where it pulls ahead.

The free window through June 22 is a good time to try it on one genuinely big task.

Fable 5's safety design routes sensitive queries instead of capping capability: cybersecurity goes to Claude Opus 4.8, biology and chemistry have dual-use protections, and distillation protections guard the model itself. Routing triggers in under 5 percent of sessions. It is free for Pro, Max, Team, and Enterprise users through June 22, 2026, then requires computing credits, with API pricing at $10 per million input tokens and $50 per million output tokens. Save it for long, complex work and let regular Claude models handle the quick stuff.