Claude Fable 5 and Mythos 5: Mythos-Class Capability Goes General, With Caveats

Anthropic launched Claude Fable 5 and Claude Mythos 5, bringing its most capable "Mythos-class" model to general use for the first time. Fable 5 is the safe-for-general-use version, available everywhere today and callable on the Claude API as claude-fable-5. Mythos 5 is the same underlying model with safeguards lifted in some areas, restricted to a small group of cyberdefenders and infrastructure providers through Project Glasswing. Both are priced at $10 per million input tokens and $50 per million output tokens -- less than half the price of Claude Mythos Preview, per Anthropic.

This is the model tier the Opus 4.7 launch was building toward: that release shipped the first cyber safeguards Anthropic said it would test on less capable models before a broad Mythos-class release. Fable 5 is that broad release. For context on the Opus line that now sits beneath it, see the Opus 4.8 Fast mode coverage -- Opus 4.8 is also the model Fable 5 falls back to when a safety classifier fires.

Key Takeaways

• Fable 5 is generally available today (June 9, 2026) and addressable on the Claude API as claude-fable-5. Anthropic calls it a Mythos-class model made safe for general use.
• Mythos 5 is the same model with safeguards lifted in some areas, restricted to vetted partners. It first deploys via Project Glasswing with cyber safeguards lifted, and soon to select biology researchers with biology/chemistry safeguards lifted.
• Pricing is $10 per million input tokens and $50 per million output tokens -- Anthropic says this is less than half the price of Claude Mythos Preview.
• Free subscription access is time-boxed. June 9-22, Fable 5 is included on Pro, Max, Team, and seat-based Enterprise plans at no extra cost. On June 23, Anthropic says it will remove Fable 5 from those plans and usage will require usage credits, until capacity allows.
• Safeguards add fallbacks. New classifiers cover cybersecurity, biology/chemistry, and distillation; some requests fall back to Claude Opus 4.8. Anthropic says classifiers trigger on average in under 5% of sessions and more than 95% of Fable sessions involve no fallback.
• Mandatory 30-day data retention applies to all traffic on Mythos-class models, on first- and third-party surfaces. Anthropic says it will not use this data to train new models or for non-safety purposes.

What Are Claude Fable 5 and Mythos 5?

Fable 5 and Mythos 5 are two deployments of the same underlying Mythos-class model. Fable 5 is the version Anthropic has made safe for general use, with safeguards in place; Mythos 5 is the same model with safeguards lifted in specific areas and access restricted to vetted partners. Anthropic positions Fable 5 as exceeding any model it has ever made generally available and as state-of-the-art on nearly all tested AI capability benchmarks.

The split matters because it separates capability from access. Fable 5 gives everyone the underlying model with guardrails; Mythos 5 gives a small set of trusted organizations the same model with certain guardrails removed for legitimate high-stakes work. Anthropic says Claude Mythos Preview users can upgrade to Claude Mythos 5.

How Fable 5 and Mythos 5 Differ

The two share one model. What changes is which safeguards are active and who can use them.

Mythos 5 initially deploys through Project Glasswing, in collaboration with the US government, as an upgrade to Claude Mythos Preview. Access is for cyberdefenders and infrastructure providers with cyber safeguards lifted, and soon for select biology researchers with biology and chemistry safeguards lifted, until broader trusted access is available.

Pricing and Subscription Access: Mind June 23

Both models cost $10 per million input tokens and $50 per million output tokens. Anthropic says that is less than half the price of Claude Mythos Preview, so teams already on the preview tier should see their per-token costs drop on the upgrade.

The subscription story is more time-sensitive and worth planning around. From today through June 22, Fable 5 is included on Pro, Max, Team, and seat-based Enterprise plans at no extra cost. On June 23, Anthropic says it will remove Fable 5 from those plans and usage will require usage credits. Anthropic says it aims to restore Fable 5 as a standard part of subscription plans when capacity allows, but gives no date. If your team plans to standardize on Fable 5 inside a subscription plan, the free window is short -- budget for usage credits after June 22 or confirm a restoration timeline before committing workflows to it.

Capability Claims

Anthropic makes strong capability claims for Fable 5, and the pattern in its framing is that the lead grows with task difficulty: the longer and more complex the task, the larger Fable 5's lead, and both Fable 5 and Mythos 5 can work autonomously longer than prior Claude models. These are Anthropic's claims; independent third-party validation across these benchmarks is not something this article can confirm.

The examples Anthropic cites span coding, knowledge work, vision, and science:

• Software engineering. Stripe reported a codebase-wide migration across a 50-million-line Ruby codebase, which Anthropic frames as compressing months of engineering into days. Cognition's FrontierCode evaluation is cited as showing Fable 5 highest among frontier models, even at medium effort.
• Knowledge work. Hebbia reported a top score on its Finance Benchmark; IMC reported strong results on trading-analysis evaluations.
• Vision. Anthropic cites scientific-figure interpretation, rebuilding a web app's source code from screenshots, and beating Pokemon FireRed with a minimal, vision-only harness.
• Long-context and memory. Anthropic highlights improved long-context and memory behavior, part of why it says the model sustains longer autonomous runs.
• Science (Mythos 5). Anthropic cites internal drug-design and molecular-biology hypothesis work, including accelerating aspects of drug design and producing candidates against multiple protein targets.

The practical read for builders: Fable 5 is pitched at the long-horizon, hand-it-off agentic work where prior models needed supervision. If your workloads are short and simple, the launch framing suggests the gap over Opus 4.8 narrows.

Safeguards, Fallbacks, and the 30-Day Retention Rule

This is the part operators need to read carefully, because it changes runtime behavior and data handling.

Fable 5 ships new safety classifiers covering cybersecurity, biology/chemistry, and distillation. When a classifier fires, some requests fall back to Claude Opus 4.8 instead of being served by Fable 5. Anthropic says the classifiers trigger on average in under 5% of sessions and that more than 95% of Fable sessions involve no fallback at all. It also cautions that the safeguards are deliberately cautious and may catch harmless requests, and that universal jailbreaks are likely impossible to completely prevent. The operational implication: a small fraction of your traffic may silently be answered by a different, less capable model, so test sensitive prompt paths and watch for quality shifts on flagged topics.

The data-retention policy is the second operational change. Anthropic will require 30-day retention for all traffic on Mythos-class models, on both first- and third-party surfaces. It says it will not use this data to train new Claude models or for non-safety-related purposes, that human access is logged, and that data is deleted after 30 days in almost all cases. For regulated or privacy-sensitive workloads, this is a material difference from standard API data handling -- review it against your compliance posture before routing production traffic through Fable 5 or Mythos 5.

Who Should Use What

Choose Fable 5 if you want the most capable generally available Claude model and can accept the new safeguards and the 30-day retention requirement. Reach for it on long-horizon agentic work -- large migrations, multi-step analysis, autonomous runs -- where Anthropic says its lead is largest. Confirm your subscription budget against the June 23 change to usage credits.

Pursue Mythos 5 only if you are an organization that qualifies for trusted access: cyberdefenders and infrastructure providers via Project Glasswing today, or select biology researchers soon. It is the same model with specific safeguards lifted, and it carries the same pricing and retention rules. For everyone else, Fable 5 is the model you can actually deploy today.

FAQ

For quick answers on what Fable 5 is, how Mythos 5 differs, pricing, and the safeguards and retention rules, see the structured FAQ in this article's metadata -- it covers the same ground as the sections above in short form.

Sources

• Anthropic, "Claude Fable 5 and Claude Mythos 5" (June 9, 2026): https://www.anthropic.com/news/claude-fable-5-mythos-5