Anthropic rolled out Claude Fable 5.2 with lower API pricing and a restricted Mythos 5.2 for vetted organizations.

Anthropic launches Claude Fable 5.2 and Mythos 5.2

Anthropic launches Claude Fable 5.2 and Mythos 5.2

Image: AI-generated editorial illustration

Anthropic announced Claude Fable 5.2 and Claude Mythos 5.2 on Nov. 9, positioning the new models for technical work while cutting API prices and expanding access across its own platform and major cloud marketplaces. Fable 5.2 is the company’s most capable model for technical work, according to Anthropic, while Mythos 5.2 is restricted to vetted organizations and aimed at cybersecurity and biology research.

Anthropic announced Claude Fable 5.2 and Claude Mythos 5.2 on Nov. 9, introducing a pair of models it says are designed for technical work, with lower API pricing for Fable and tighter access rules for Mythos.

The company describes Claude Fable 5.2 as its most capable model for technical work and says it sets a new bar in coding, finance and the sciences. Anthropic also says Claude Mythos 5.2 is its most capable model class for cybersecurity and biology research. In Anthropic’s documentation, Mythos 5.2 is the same underlying model as Fable 5.2, but it is restricted to organizations verified through Anthropic’s verification programs.

What changed in Anthropic’s model lineup?

Anthropic split the launch into two product names aimed at different customer groups. Fable 5.2 is broadly available to Pro, Max, Ultra, Team and Enterprise users, while developers can access it through the Claude Platform and listed cloud marketplaces. Anthropic says Fable 5.2 is also available natively on the Claude Platform and through AWS, Google Cloud and Microsoft Foundry.

Mythos 5.2, by contrast, is limited to vetted organizations. Anthropic says that restriction is tied to its verification programs, and its docs frame Mythos as the same base model as Fable with a narrower release path for sensitive workloads.

Anthropic’s documentation says both models support text-and-image input and text output, have a 1 million-token context window, can produce up to 128,000 tokens of output, and use always-on adaptive thinking. The company also says 1 million tokens is roughly 555,000 words or 2.5 million Unicode characters on its current tokenizer.

How does the pricing compare with earlier versions?

Anthropic lowered Fable 5.2’s standard API pricing to $8 per million input tokens and $40 per million output tokens, down from $10 and $50 for Fable 5.1. Cache reads are priced at $0.20 per million tokens.

The company also lists a fast mode for Fable 5.2 on the Claude Platform, Google Cloud, Microsoft Foundry and Claude Code. Anthropic says fast mode can run up to 2.5 times faster and is priced at $16 per million input tokens and $80 per million output tokens. For batch processing, Anthropic lists both Fable 5.2 and Mythos 5.2 at $4 per million input tokens and $20 per million output tokens.

Mythos 5.2 starts at the same base prices as Fable’s standard tier: $8 per million input tokens and $40 per million output tokens. Anthropic also says prompt cache writes for Mythos 5.2 cost $10 per million tokens for 5 minutes and $16 per million tokens for 1 hour.

What does Anthropic say the models can do?

Anthropic says Fable 5.2 can handle ambiguous problems for days, verify its own work and orchestrate subagents in a loop. In testing, the company says the model can run ambiguous engineering work for days and coordinate hundreds or more than a thousand subagents.

Anthropic also released company-reported benchmark figures for Mythos 5.2, including 49.1% on FrontierCode 2.0, 74.6% on Terminal-Bench-Science 0.1, 89.3% on OSWorld 2.1, 89.6% on CVE Rediscovery and 85.1% on a 24-hour de novo binder-design evaluation. Those results are Anthropic’s own claims and were not independently verified in the material provided.

The company says Mythos 5.2 is its most capable model class for cybersecurity and biology research, and that Claude Security now runs on Mythos 5.2.

What guardrails and platform limits come with the launch?

Anthropic says many cybersecurity and biology queries on Fable 5.2 are automatically routed to less capable models when safeguards flag them. Most Claude applications route flagged cybersecurity queries to Opus 4.8 and flagged biology queries to Opus 5, according to Anthropic.

The company says teams affected by those safeguards can apply to verification programs for adjusted settings. Anthropic also says Enterprise Frontier Safeguards stores prompts and responses in a customer’s infrastructure to support automated safety monitoring that is compatible with zero data retention, while eligible Cyber Verification Program customers that already have zero data retention can keep it for Mythos 5.2 until EFS is available.

Anthropic says both models require 30-day data retention by default for safety monitoring and are not available under zero data retention unless expressly authorized by Anthropic. Its docs also say Fable 5.2 and Mythos 5.2 enforce prefix checks for every account, regardless of age.

The models also have a few platform-specific limits. Anthropic says Fable 5.2 and Mythos 5.2 reject the older computer_20251124 computer-use tool on every platform except Amazon Bedrock, and only Fable 5.2 and Mythos 5.2 read each other’s thinking blocks; other models do not. Moving from either model to another model loses the reasoning from those turns, Anthropic says. For refused Fable 5.2 requests, permitted fallback targets are Claude Opus 4.8 and Claude Opus 5.

Mythos 5.2’s model overview also says it will not retire sooner than Nov. 9, 2027, and sets a reliable knowledge cutoff of June 2026.

Sources