Claude Fable 5.2 and Mythos 5.2 launch with lower API prices

Anthropic launched Claude Fable 5.2 and Claude Mythos 5.2 on November 9, with lower API prices and an optional faster mode. Fable is the broadly available version. Mythos uses the same underlying model with different cybersecurity and biology safeguards, and access remains subject to verification.

For anyone choosing between them, access is the first question to settle. A published price tells you what usage costs. It does not establish that your account can use every model or that an existing application can switch without changes.

What Claude Fable 5.2 costs

The launch announcement prices both models at $8 per million input tokens and $40 per million output tokens. Optional fast mode doubles those rates to $16 and $80. Anthropic says normal output generation is up to 30% faster than version 5.1, while fast mode can increase generation speed by up to 2.5 times.

The Fable 5.2 documentation lists a one million token context window and a 128,000 token output ceiling. Input supports text and images, with text output. Cache reads cost $0.20 per million tokens, while the Batch API offers a 50% discount on input and output. Mythos shares Fable’s listed specifications and pricing.

Compare the cost of a completed task, including retries and review time, before moving a whole workflow. A faster answer can be useful, but paying twice the token rate needs a concrete benefit when the task already runs comfortably in the background.

Where access is available

Anthropic says Fable 5.2 is available across its platforms and cloud partners, including AWS, Google Cloud and Microsoft Azure. Mythos requires access through the Cyber or Life Sciences Verification Programs. Enterprise customers can also encounter Mythos through Claude Security.

The model documentation is more specific about deployment routes. It lists Claude API, Claude Platform on AWS and both Google managed and Anthropic managed Google Cloud offerings, using claude-fable-5-2. It does not list an Amazon Bedrock or Microsoft Foundry identifier. Check the route your application actually uses before treating the broad launch announcement as an endpoint migration guide.

Subscription capacity is another separate check. Our guide to Claude Ultra and Max usage limits explains the earlier plan changes. It should not be read as a new Fable 5.2 entitlement or a promise of unlimited use.

Fable launch offers and reset deadlines

Anthropic’s launch offer adds temporary subscriber benefits from November 9. Pro and Team Standard receive a week of Fable access, capped at 50% of weekly usage, plus one reset. Both expire November 16, 2026 at 11:59 p.m. PT. Further Fable use returns to usage credits if enabled.

Max, Ultra, Team Premium and legacy seat based Enterprise Premium receive one reset for subscriptions under six months, two for six to 12 months, or three beyond 12 months. These expire March 8, 2027 at 11:59 p.m. PT. Tenure runs from the first paid bill, including for organizations.

Reset eligibility requires paid billing before November 9 at 10 a.m. PT, plus earlier organization membership for Team and Enterprise seats. Trial exceptions apply. Free, other Enterprise tiers and HIPAA ready Enterprise plans are excluded. Check the offer’s exact conditions against your seat type.

These are banked resets that you redeem, not a universal automatic refill. Anthropic’s reset instructions say a Fable reset restores weekly usage spent on Fable while leaving other models’ weekly usage unchanged. If a five hour limit applies, that whole window refills. Limits do not increase, and billed extra usage is not refunded.

Redeem in Settings > Usage on Claude web, desktop or mobile. The restored allowance also applies in Claude Code, though its terminal and IDE do not provide the redemption button. Redemption cannot be undone. Downgrading or cancelling removes unused resets.

What Mythos verification means

Anthropic’s Cyber Verification Program guidance says organizations apply once, with administrators allocating access. Independent researchers can apply individually, but only for the Defense tier. Anthropic aims to respond within seven business days, which is a review target rather than guaranteed access.

Approval also needs to be provisioned to the appropriate workspace, role or linked account. The help page distinguishes work on an organization’s own systems from services delivered to customers, which require separate approval. It says managed service applications are open, while applications for customer operated products are not yet open.

That page still names Mythos 5.1 in several sections. Use the November 9 launch for the new model announcement and confirm your actual grant before assuming older plan allowances or provider timing apply unchanged to Mythos 5.2.

What to check before switching

Anthropic reports stronger coding and reasoning results, but its launch benchmarks remain vendor presented evidence. They are a reason to evaluate the model against your own difficult tasks, not a guarantee that it will improve every job.

For developers, adaptive thinking is always enabled. The model page also warns that forced tool selection can fail, and that 30 day data retention applies unless Anthropic expressly authorizes an exception. Those requirements deserve review before a production switch.

Start with a bounded trial: representative inputs, clear success criteria and a spending limit. Keep the old workflow available until the new model has shown that its finished work justifies the change. That is a more useful test than choosing solely from the largest benchmark headline.