Is Claude Fable 5 the same model as Claude Mythos 5?

They share the same underlying architecture and capabilities. The difference is safety filters: Fable 5 can decline requests that trip its cybersecurity, biology, or distillation classifiers, while Claude Mythos 5 — available only to vetted partners — does not.

How much does Claude Fable 5 cost?

Through the API it’s $10 per million input tokens and $50 per million output tokens, as of July 2026, per Anthropic’s model documentation . On Claude.ai it’s included in the Pro plan (from about $17/month) and Max plan (from $100/month) rather than billed separately.

Why does Claude Fable 5 sometimes refuse a request?

One of three built-in classifiers — covering offensive cybersecurity, dual-use biology and chemistry, or attempts to extract its capabilities into another model — has flagged the request. Anthropic says this happens in under 5% of conversations, and the API can automatically retry a declined request on another Claude model.

Is Claude Fable 5 stronger than Claude Opus 4.8?

Yes — Anthropic positions Fable 5 as a significant step up from Opus 4.8 in reasoning, coding, and long-running agentic tasks, while keeping comparable safety alignment. Declined Fable 5 requests can fall back to Opus 4.8 automatically.

Can I use Claude Fable 5 on the free version of Claude?

No — it’s available to Pro, Max, Team, and Enterprise subscribers on Claude.ai, and to any developer with API access, but not on the free tier.

What Is Claude Fable 5? Anthropic's Flagship AI Model, Explained

Claude Fable 5 is Anthropic’s most capable AI model available to the public — a system built for demanding reasoning, coding, and long-running autonomous work. It ships with built-in safety filters that its internal-only sibling, Claude Mythos 5, does not have.

What it is

Claude Fable 5 (API model ID claude-fable-5) is the flagship model in Anthropic’s Claude family. It became generally available on June 9, 2026, through Claude’s chat app, the Claude API, and major cloud platforms including Amazon Bedrock, Google Cloud, and Microsoft Foundry. On Claude.ai, it’s available to Pro, Max, Team, and Enterprise subscribers — it isn’t part of the free tier.

The name comes from the Latin fabula, “that which is told” — the same root idea as the Greek mythos, which names its sibling model. Anthropic built both on the same underlying architecture; by the company’s own account, the only real difference between them is which safety filters are switched on.

How it differs from Claude Mythos 5

Claude Mythos 5 shares Fable 5’s full capabilities but ships without the classifiers that can make Fable 5 decline a request. Mythos 5 isn’t generally available — it’s restricted to vetted cybersecurity firms and biology researchers approved through Anthropic’s Project Glasswing program. Everyone else gets essentially the same underlying model through Fable 5, just with guardrails attached.

What it can do

Fable 5 is built to handle long, multi-step tasks that earlier models couldn’t sustain on their own. Anthropic says that, run inside an agent tool like Claude Code, it has completed month-long software projects in a matter of days and compressed a 50-million-line codebase migration into a single day. It carries a 1-million-token context window by default, can reason over charts, tables, and scanned documents with high precision, and keeps “adaptive thinking” switched on at all times — automatically scaling how much it reasons to how hard the question is, rather than requiring a user to pick a mode.

Why the safety filters matter

Three automated systems can make Fable 5 decline a request outright: one blocks offensive cybersecurity queries, one restricts dual-use biology and chemistry research, and one is designed to stop people from using Fable 5’s own answers to train a rival model — a practice known as model distillation. Anthropic says fewer than 5% of conversations trigger any filter, and independent red-teaming found no way to bypass all three at once across more than 1,000 hours of testing. When a request is declined, the API can automatically retry it on another Claude model rather than simply failing outright.

These aren’t abstract precautions. In June 2026, U.S. export-control rules briefly restricted Fable 5 and Mythos 5 access worldwide after researchers found a way to prompt around the cybersecurity filter — a reminder of how much weight regulators place on these safeguards for a model this capable.

What it costs

Through the API, Fable 5 costs $10 per million input tokens and $50 per million output tokens, as of July 2026, per Anthropic’s official model documentation — with the usual discount for cached prompts. On Claude.ai, it isn’t billed separately: it’s bundled into the Pro plan (from about $17–20/month) and the Max plan (from $100/month).

In the news

The distillation filter isn’t a hypothetical safeguard. The White House has accused Moonshot AI of using Fable 5’s outputs to train its own model — precisely the scenario that filter is built to catch.