Reviewed by Jonathan West · Updated Nov 9, 2026

Claude Mythos 5.2 Explained

A complete operational guide to verified access, benchmark results, pricing, and system changes for Anthropic's restricted frontier model.

Reviewed by Jonathan West · Updated Nov 9, 2026

Claude Mythos 5.2 is Anthropic's restricted-access frontier model, launched on November 9, 2026. It shares the same underlying model as Claude Fable 5.2 and is available only to verified organizations.

Anthropic prices Claude Mythos 5.2 at $8 per million input tokens and $40 per million output tokens, matching Fable 5.2 and cutting rates 20 percent compared to Claude Mythos 5.1. While standard users access Fable 5.2 with strong cybersecurity and biology safeguards, Mythos 5.2 offers the same capabilities to organizations verified through dedicated verification programs.


What Claude Mythos 5.2 Is and Who Can Get Access

Claude Mythos 5.2 is Anthropic's restricted-access frontier model, launched on November 9, 2026 under the API identifier claude-mythos-5-2. Anthropic built Mythos 5.2 on the identical underlying architecture as Claude Fable 5.2, but gates access behind strict institutional vetting rather than making it publicly available.

Access is restricted to organizations approved through Anthropic verification programs. Cybersecurity professionals must apply through the Cyber Verification Program, while life sciences researchers apply through the Life Sciences Verification Program. Enterprise customers can also access the underlying technology via Claude Security, which runs across the model family.

Anthropic lists Claude Mythos 5.2 on its Covered Models documentation under a designation of limited availability for approved partners. See Claude Fable 5.2 vs Mythos 5.2 for access routing.


Claude Mythos 5.2 Facts at a Glance

Claude Mythos 5.2 provides a standard 1M-token context window alongside a 128,000 maximum output token limit. The architecture matches its public sibling, detailed in the Claude Fable 5.2 explained overview, while enforcing distinct access boundaries.

SpecificationValueSource
API Model Identifierclaude-mythos-5-2Anthropic Release Notes
Access StatusLimited availability (approved partners)Anthropic Covered Models
Context Window1M-token context window (default and maximum)Anthropic Models Overview
Maximum Output Tokens128,000 tokensAnthropic Whats New in Fable 5.2
Platform DeploymentClaude API for verified accountsAnthropic Announcement
Earliest Retirement DateNot sooner than November 9, 2027Anthropic Model Deprecations
Regulatory DesignationCovered Model (30-day minimum retention)Anthropic Covered Models

Both models run adaptive thinking by default at high effort. They use the tokenizer introduced with Claude Opus 4.7. Their reliable knowledge cutoff is June 2026. The full 1M-token context window is billed at standard per-token pricing with no long-context surcharges.


Claude Mythos 5.2 Pricing and Access

Claude Mythos 5.2 matches the baseline rate card of Claude Fable 5.2, billing $8 per million input tokens and $40 per million output tokens on the Claude API. Complete tier structures and cache costs are detailed in the guide to Claude Fable 5.2 pricing.

On the Anthropic pricing schedule, a 5-minute cache write costs $10 per million tokens, a 1-hour cache write costs $16 per million tokens. Cache reads cost $0.20 per million tokens (0.025 times the base input rate). Asynchronous requests through the Batch API run at a 50 percent discount, priced at $4 per million input tokens and $20 per million output tokens.

Anthropic offers an optional fast mode for verified Mythos organizations that generates up to 2.5 times higher output tokens per second, priced at $16 per million input tokens and $80 per million output tokens. Fast mode cannot combine with the Batch API. It stacks with prompt caching and the 1.1 times multiplier for United States data residency. See Claude Fable 5.2 vs Mythos 5.2 for the application workflows.


What Anthropic Reports for Mythos 5.2

Anthropic reported specialized domain evaluation results for Claude Mythos 5.2 across cybersecurity scenarios, common vulnerabilities and exposures rediscovery, and molecular biology workflows. These domain benchmarks compare Mythos 5.2 directly against Mythos 5.1 and Claude Opus 5.5 without safeguards.

BenchmarkClaude Mythos 5.2Claude Mythos 5.1Claude Opus 5.5 (Without Safeguards)
CyScenarioBench (dagger)76.3%61.7%67.6%
CVE Rediscovery89.6%78.3%81.3%
De novo binder design (24 hr)85.1%79.1%82.6%

On CyScenarioBench, each lab on unique tasks, Anthropic reported 76.3% accuracy for Mythos 5.2 compared to 61.7% for Mythos 5.1, 67.6% for Opus 5.5 without safeguards, and 61% for GPT-6 Cyber. On CVE Rediscovery, Mythos 5.2 reached 89.6%. Opus 5.5 without safeguards scored 81.3% and Mythos 5.1 scored 78.3%. On 24-hour de novo binder design, Mythos 5.2 achieved 85.1%, outperforming Opus 5.5 without safeguards at 82.6% and Mythos 5.1 at 79.1%.


What Changed From Claude Mythos 5.1

Claude Mythos 5.2 lowers operational token pricing by 20 percent across all standard billing units compared to its predecessor. Internal cost modeling from Anthropic's published rates confirms that base input drops from $10 to $8 per million tokens, base output drops from $50 to $40, 5-minute cache writes drop from $12.50 to $10, 1-hour cache writes drop from $20 to $16. Cache reads drop from $0.25 to $0.20 per million tokens.

Anthropic states that Claude Mythos 5.2 generates outputs up to 30 percent faster than Claude Mythos 5.1 and Claude Fable 5.1. Our calculation from Anthropic's table puts the Mythos 5.2 gains over Mythos 5.1 at 14.6 points on CyScenarioBench, 11.3 on CVE Rediscovery and 6.0 on de novo binder design.

Anthropic lists four changes from Claude Fable 5.1 that can break code on Claude Fable 5.2. The prefix check applies to every account. Claude Fable 5.1 cannot read Claude Fable 5.2 thinking blocks. The earlier computer_20251124 tool returns an error on every platform except Amazon Bedrock. With the advisor tool, a Claude Fable 5.2 executor needs a Claude Fable 5.2 or Claude Mythos 5.2 advisor. This page does not verify which of these apply to Claude Mythos 5.2 integrations, so check them against Anthropic's migration guide.

  • CyScenarioBench: Mythos 5.2 leads Mythos 5.1 by 14.6 percentage points (a 23.7 percent relative improvement).
  • CVE Rediscovery: Mythos 5.2 leads Mythos 5.1 by 11.3 percentage points (a 14.4 percent relative improvement).
  • De novo binder design: Mythos 5.2 leads Mythos 5.1 by 6.0 percentage points (a 7.6 percent relative improvement).
  • Terminal-Bench-Science 0.1: Mythos 5.2 scored 75.3%, outperforming public Fable 5.2 at 74.6% by a 0.7 percentage point gap.

How Long It Stays Available

Anthropic lists the retirement status of Claude Mythos 5.2 as active, with a tentative retirement date set to not sooner than November 9, 2027. That floor is one calendar year after launch.

Under Anthropic's deprecation policy, Anthropic commits to providing at least 60 days of advance notice before retiring a publicly released model. The earlier Claude Mythos 5.1 release carries a tentative retirement floor of not sooner than September 1, 2027. Mythos 5 is listed as not sooner than June 9, 2027.

Organizations operating on Mythos 5.1 should review their tool integrations and prompt prefixes against the newer API specifications before transitioning workloads to Claude Mythos 5.2.

Frequently Asked Questions

  • Claude Mythos is Anthropic's restricted-access release of the same underlying model as Claude Fable, available only to organizations verified through Anthropic's verification programs.
  • Yes, Claude Mythos 5.2 and Claude Fable 5.2 share the exact same underlying model architecture. The distinction lies in access controls and safeguards: Fable 5.2 has strong cybersecurity and biology safeguards making it available to all users, whereas Mythos 5.2 is restricted to organizations verified through Anthropic's verification programs.
  • Claude Mythos 5.2 costs $8 per million input tokens and $40 per million output tokens on the Claude API, matching Fable 5.2. Prompt cache writes cost $10 per million tokens (5-minute) or $16 per million tokens (1-hour), cache reads cost $0.20 per million tokens, and the Batch API provides a 50 percent discount at $4 input and $20 output.
  • Token rates dropped 20 percent across all units, output speed increased up to 30 percent, and reported benchmark scores rose by 14.6 points on CyScenarioBench and 11.3 points on CVE Rediscovery. Developer changes include updating to the computer_toolset_20260801 toolset and advisor pairing rules.
  • Claude Mythos 5.2 was launched on November 9, 2026 alongside Claude Fable 5.2.
  • Organizations verified through Anthropic's verification programs, such as the Cyber Verification Program and the Life Sciences Verification Program, can use Claude Mythos 5.2. Everyone else can use Claude Fable 5.2, which is the same underlying model with stronger safeguards.

Plan Your Claude Mythos Architecture

Layer3Labs helps engineering teams review model integrations, verify compliance controls, and optimize multi-agent API costs. Book a consultation to assess whether Claude Mythos 5.2 or public frontier models fit your security architecture.

Book a Consultation