What Is Claude Fable 5? Anthropic’s New Model, Explained

On June 9, 2026, Anthropic released Claude Fable 5 — and for once, the launch-day superlatives are doing real work. Fable 5 is the public version of Mythos, the internal frontier model Anthropic unveiled in April 2026 and deliberately kept on a short leash. Two months later, the company says it has found a way to ship Mythos-class capability that is, in its words, safe for general use.

If you’ve seen “Fable 5” trending and wondered whether it’s a model, a product, or a codename — here’s the whole picture in plain English.

From Mythos to Fable

The unusual part of this launch is the two-step release. Anthropic announced Mythos in April 2026 but limited who could use it, citing the model’s advanced capabilities in sensitive areas — most notably cybersecurity. That caution drew attention: a frontier lab publicly saying a model was, for the moment, too capable to release broadly.

Fable 5 is the answer to that problem. Rather than waiting for the risk landscape to change, Anthropic wrapped Mythos-class capability in a new safeguard system and shipped the combination under a new name. The naming is almost on the nose: a myth retold safely becomes a fable.

What it’s good at

Anthropic says Fable 5’s capabilities exceed those of any model it has ever made generally available, with standout performance in:

• Software engineering — writing, debugging, and reviewing real codebases, not just toy snippets. It went live in GitHub Copilot on launch day, which tells you who Anthropic expects to use it first.
• Knowledge work — research, analysis, long-document reasoning, and writing that holds up under expert review.
• Vision — reading screenshots, charts, documents, and interfaces accurately.
• Scientific reasoning — multi-step technical problem-solving across domains.

That combination matters for one use case above all: agentic AI. Agents live or die by their ability to plan multi-step work, use tools correctly, and notice when something went wrong. Each generation of frontier models has made agents noticeably more reliable, and Fable 5 is squarely aimed at that frontier.

The safety design (the interesting part)

Fable 5’s release mechanism is genuinely novel. Instead of blunting the whole model, Anthropic added targeted safeguards for specific high-risk areas — primarily advanced cybersecurity and biology:

• For a narrow set of dangerous requests, the model blocks the response outright.
• For some sensitive topics, queries are rerouted to Claude Opus 4.8, the previous flagship, rather than answered by Fable 5 itself.
• Anthropic says this routing triggers in fewer than 5% of sessions on average — most users will never notice it.

You can read this two ways, and both are true: it’s a real safety mechanism, and it’s a template for how labs can ship increasingly capable models without an all-or-nothing release decision.

Pricing and availability

Pricing and availability are as reported at launch and may change — Anthropic’s announcement is the authoritative source.

Why this matters for AI agents

A chatbot answers; an agent finishes. The difference shows up in the boring middle of a task: step seven of eleven, where a weaker model loses the plot, calls the wrong tool, or confidently declares victory early. Frontier-model upgrades disproportionately improve exactly that middle stretch — long-horizon reliability — which is why agent products tend to improve in lockstep with model releases.

Practically, models like Fable 5 mean agents that can hold a longer plan in their head, recover from their own mistakes, and collaborate usefully in multi-agent teams, where one agent’s sloppy output becomes another agent’s garbage input. Better models raise the ceiling for every agent stacked on top of them.

How to try Fable 5 today

1. Claude apps — paid subscribers can select Fable 5 directly in Claude’s products.
2. GitHub Copilot — generally available since June 9 for Copilot users.
3. The API — model ID claude-fable-5 for developers building their own tools.
4. Agentic AI on iOS — our app runs on frontier models including Claude Fable 5, wrapped in ready-made expert agents, with no setup and no account.

Quick answers

Is Fable 5 the same as Mythos? It’s the same class of capability, packaged with safeguards for general release. Anthropic describes Fable 5 as the version of Mythos that’s safe to ship broadly.

Is it free? No — it’s available to paid subscribers and API customers. Apps built on the API (like ours) can offer free tiers on top.

What happens when I hit a “blocked” topic? In under 5% of sessions, sensitive queries get rerouted to Opus 4.8 or declined. For everyday work — writing, code, research, planning — you’ll simply never see it.

Sources: Anthropic — Claude Fable 5 and Claude Mythos 5 · CNBC · TechCrunch · 9to5Google · GitHub Changelog