Claude Mythos 5.2

Claude Mythos 5.2 is Anthropic's verification-only deployment of the same underlying model as Claude Fable 5.2, with more permissive safeguards in biology and cybersecurity, the high-risk dual-use domains where Fable carries additional safeguards. Access is limited to organizations verified through Anthropic's Cyber Verification Program or Life Sciences Verification Program; Claude Enterprise customers can also reach it indirectly through Claude Security. It shares Fable 5.2's specifications: text and image input, text output, tool use, adaptive thinking that is always on (default effort high), a 1M-token context window, 128K max output, and June 2026 knowledge and training-data cutoffs. Pricing matches Fable 5.2 at $8/$40 per million input/output tokens, with cache reads at $0.20, 5-minute cache writes at $10 and 1-hour cache writes at $16; fast mode ($16/$80) is limited to verification-program members and is not offered on Amazon Bedrock. Anthropic calls it its most capable model for cybersecurity engineering, rediscovering 89.6% of 1,829 October 2026 CVEs versus 78.3% for Mythos 5.1. Self-reported scores (safeguards off) include Terminal-Bench 4.0 70.5%, Terminal-Bench-Science 0.1 75.3% and de novo protein binder design 85.1%. Available as `claude-mythos-5-2` on the Claude API, Claude Models on Google Cloud and Microsoft Foundry, and as `anthropic.claude-mythos-5-2` on Amazon Bedrock, where the Cyber Verification Program is currently open only to organizations with a data retention exemption for Fable 5.2 or Fable 5.1.