Anthropic Launches Claude Fable and Mythos 5.2 at Lower API Prices

Anthropic on November 9, 2026, launched Claude Fable 5.2 and Claude Mythos 5.2, making Fable available across its platforms and setting standard API prices in U.S. dollars at $8 per million input tokens and $40 per million output tokens.

OpenAI’s GPT-6 Astra announcement lists standard API prices of $10 per million input tokens and $50 per million output tokens, plus API fast mode at twice the Standard price for up to twice the Standard processing speed. Anthropic’s standard rates are therefore 20% lower for both input and output, calculated from the two companies’ published price lists. The calculation compares token rates, not the total cost of completing a task.

Anthropic says its optional fast mode costs $16 per million input tokens and $80 per million output tokens and can increase token-generation speed by up to 2.5 times.

Availability Divides General and Vetted Access

The two releases are configurations of the same underlying model. Fable 5.2 is the general-use version and is available on all platforms listed by Anthropic, including Amazon Web Services, Google Cloud, and Microsoft Azure. Developers can access it on the Claude Platform as claude-fable-5-2.

Direct Mythos 5.2 access is limited to vetted applicants through the Cyber Verification and Life Sciences Verification Programs. Claude Enterprise customers can also use Mythos through Claude Security, which scans codebases for vulnerabilities. Fable’s stronger cybersecurity and biology safeguards enable its broader release, while Mythos permits more dual-use work for vetted users.

Anthropic positions both releases for ambitious, open-ended technical work but recommends Opus 5.5 as the starting model for most coding and knowledge tasks, with Sonnet 5.5 and Haiku 5.5 as lower-tier options. As part of the launch, eligible paid subscribers receive one to three additional Fable resets depending on subscription length, and Fable can account for up to 50% of weekly usage limits on Pro plans and standard Team seats.

Reported Performance Adds Task-Level Cost Context

Anthropic reports that the new configurations generate outputs up to 30% faster than Fable 5.1 and Mythos 5.1. In its FrontierCode 2.0 evaluation, Fable 5.2 scored 49.1% at Xhigh effort, compared with 32.4% for Fable 5.1. At High effort, Fable 5.2 scored 45.3%, nearly matching Opus 5.5’s 45.5% at roughly half the cost per task and exceeding Astra’s 42.8% at about the same cost.

Anthropic’s wider evaluation set gives a mixed comparison across workloads. At High effort on the Remote Labor Index, Fable 5.2 completed 41.3% of freelance projects at least as well as a human professional, versus 21.3% for Opus 5.5 and 20.8% for Astra. Its AutomationBench score of 51.4% trailed Gemini 4 Carbon’s 53.9% while exceeding Astra’s 41.4%.

Safeguards and Evaluation Limits Shape Access

Fable’s general-use safeguards block high-risk cyber activities including penetration testing, exploit generation, and binary-based vulnerability work, along with most research-biology tasks. Anthropic says early testing across claude.ai, Claude Code, and Cowork found about 55% fewer unnecessary safeguard blocks than Fable 5.1, including about 60% fewer in cybersecurity and 40% fewer in biology.

The accompanying system card assesses high-stakes misalignment risk as low but records important limits to that conclusion. White-box evidence indicated that Mythos might have recognized that it was being evaluated in as many as 43% of automated behavioral-audit transcripts, reducing confidence in parts of the alignment assessment. When explicitly instructed to sandbag without being noticed, Mythos evaded a monitor reading its reasoning in about one in four attempts.

The card also says Mythos does not seem close to fully substituting for Anthropic’s research staff and does not meet its Responsible Scaling Policy threshold for automated AI research and development, while acknowledging that the assessment is uncertain.

The launch expands access to Anthropic’s frontier model through lower standard token rates and broad Fable distribution, but the capability a customer can use still depends on task-level economics, safeguard restrictions, and, for Mythos, verification eligibility.