Overview
Claude Mythos 5.2 is the restricted-access configuration of the model Anthropic also released as Claude Fable 5.2 on November 9, 2026. Mythos uses more permissive safeguards for vetted cybersecurity and life-sciences work. The first decision is whether you need direct model access for an approved workload or repository scanning through Claude Security.
What's New
Anthropic's CVE Rediscovery evaluation reports 89.6% for Mythos 5.2 versus 78.3% for Mythos 5.1. The evaluation uses repository snapshots with previously reported vulnerabilities. For a defensive scanning pilot, measure confirmed findings and triage time on repositories you are authorized to assess.
Anthropic also added Mythos 5.2 to Claude Security and made Fast Mode available on supported surfaces for organizations in its verification programs. An Enterprise security team can use Claude Security to scan repositories without joining the Cyber Verification Program. Direct Mythos model access has its own eligibility and activation requirements.
Availability & Access
| What you need | Route |
|---|---|
| Direct model access for vetted cybersecurity work | Apply through the Cyber Verification Program for the appropriate tier. After approval, activate access in an eligible workspace or linked cloud account. |
| Direct model access for vetted life-sciences research | Follow the Life Sciences Verification Program link in the launch announcement. Use the approved organization and workspace once access is provisioned. |
| Repository scanning in Claude Enterprise | Check Claude Security availability with your Enterprise administrator. Claude Security can use Mythos 5.2 to scan codebases. |
Check access in the exact environment you intend to use before migrating. Anthropic's verification guide lists separate steps for Console workspaces and linked cloud accounts, including data-sharing prerequisites on some surfaces. Review retention requirements with your administrator. The migration guide specifies 30-day retention unless Anthropic expressly authorizes zero data retention.
Compared With Previous Version
- Standard API rates: $8 input and $40 output per million tokens, down from $10 and $50 on Mythos 5.1. Cache reads fell from $0.25 to $0.20 per million. Fast Mode on the Claude API costs $16 input and $80 output per million tokens.
- Capacity: The 5.2 model ID is
claude-mythos-5-2. The API limit is shared across Mythos 5, 5.1 and 5.2. Plan a parallel pilot within the existing limit. - Session compatibility: On the Claude API and Google-managed Claude Models on Google Cloud, Mythos 5.2 can carry forward thinking blocks produced by 5.1. Switching back to 5.1 drops the blocks produced by 5.2. The beta advisor tool accepts Fable 5.2 or Mythos 5.2 as advisors to a Mythos 5.2 executor. An earlier advisor model returns HTTP 400. Start a new conversation when changing the advisor definition.
Who Should Upgrade / Who Should Wait
If your organization already has Mythos 5.1, confirm the new model is provisioned and compare both versions against the same authorized jobs. Check confirmed findings, review effort, token costs and how a rollback affects session state. Start a fresh test session if your tools or advisor settings change.
Test any integration that edits earlier messages, system instructions or tool definitions. On Amazon Bedrock, Mythos 5.2 can return HTTP 400 for edited history that Mythos 5.1 accepted. Outside Bedrock, migrate the legacy computer_20251124 tool to computer_toolset_20260801 before switching.
If you're waiting for verification, evaluate repository scanning through Claude Security if your Enterprise plan includes it. For general coding or research, start with a broadly available Claude model and check whether you need the specialized access that Mythos provides.
