Robots Atlas>ROBOTS ATLAS
Claude Mythos 5.2

Claude Mythos 5.2

Mythos 5.2 (claude-mythos-5-2) · Family: Claude
Anthropic's Claude-family model (Mythos 5.2) for cybersecurity and biology: the same model as Fable 5.2 without certain safeguards, available only to verified organizations.
✓ Active⏳ Limited accessLLMMultimodalReasoning modelTool-using modelSpecialized AIScientific AI📁 Claude
Context window
1M
tokens
Parameters
Nieujawnione publicznie
parameters
Max output
128,000
tokens
Release date
9 November 2026
Access:APIHostedDeployment:☁ Cloud

Overview

Claude Mythos 5.2 is a model in Anthropic's Claude family, released on November 9, 2026 alongside Claude Fable 5.2. It is the same underlying model as Fable 5.2 but without certain cybersecurity and biology safeguards. Anthropic describes it as its most capable model for cybersecurity engineering and life-sciences research.

Access is limited to organizations verified through Anthropic's programs — the Cyber Verification Program (security teams) and the Life Sciences Verification Program (life-sciences researchers). Mythos 5.2 also powers the Claude Security product. By default a 30-day data retention policy applies for safety monitoring.

Specifications and pricing are identical to Fable 5.2: a 1M-token context window, up to 128K output tokens, text and image input, text output. Adaptive thinking is always on (default effort: high), and the knowledge cutoff is June 2026. Pricing is $8 per million input tokens and $40 per million output tokens. The model is available via the Claude API and on Amazon Bedrock, Google Cloud, and Microsoft Foundry (ID: claude-mythos-5-2).

On Anthropic's published benchmarks, Mythos 5.2 scores 89.6% on CVE Rediscovery, 76.3% on CyScenarioBench, 85.1% on de novo binder design (24 hr), and 75.3% on Terminal-Bench-Science (versus 74.6% for Fable 5.2).

Classification
LLMMultimodalReasoning modelTool-using modelSpecialized AIScientific AI
Family: Claude
Access & deployment
APIHosted
Cloud
Weights: Closed
Key parameters
📏 Context: 1M
🧩 Parameters: Nieujawnione publicznie
✓ Tools
📥 Input: text, image, documents

Technical specification

Context window
1M
tokens
Parameters
Nieujawnione publicznie
parameters
Max output tokens
128,000
tokens per response
Knowledge cutoff
1 Jun 2026
Knowledge boundary
License
Proprietary (Anthropic) — dostęp dla zweryfikowanych organizacji (Cyber / Life Sciences Verification Programs); domyślnie 30-dniowa retencja danych dla safety monitoring
Hardware requirements
Not applicable — API access. Hosted on Anthropic and cloud-partner infrastructure.
Features:✓ Tool use
Modalities
⬇ Input
textimagedocuments
⬆ Output
textcode

Capabilities and applications

Native model capabilities
Reasoning
The model's ability to reason logically and solve complex problems.
Category: reasoning
Advanced reasoning
The ability to perform multi-step, structured reasoning: analysing problems, planning steps, and drawing conclusions from hypotheses. Reasoning-first models (e.g. GPT-5.1 Thinking) dedicate a portion of inference to chains of thought before responding.
Category: reasoning
Multi-step reasoning
Carrying out multi-step chains of reasoning across long, complex tasks.
Category: reasoning
Extended thinking mode
A reasoning-model variant with a larger inference budget: more thinking cycles, higher answer precision at the cost of response time. Choice between 'standard' and 'extended' thinking is left to the user (e.g. the selector in GPT-5.2 Pro).
Category: reasoning
Long context
Support for large context windows — tens to hundreds of thousands (or millions) of input tokens. Enables analysis of entire codebases, long documents, and many parallel conversations without losing earlier information. GPT-5.1 supports 400,000 tokens.
Category: language
Coding
Generating, analysing and modifying code in many programming languages. Covers writing functions, debugging, refactoring, code review, and creating tests. Measured by benchmarks such as HumanEval and SWE-bench.
Category: coding
Agentic capability
The model's ability to autonomously plan and execute multi-step tasks by sequentially using tools, maintaining context, and adapting to intermediate results.
Category: planning
Computer use
The model's ability to operate a computer interface by interpreting screenshots and generating actions such as clicks, typing, and navigating applications.
Category: planning
Structured output
Producing data in structured formats such as JSON.
Category: structured_generation
Image understanding
Analysing and interpreting the content of images.
Category: vision
Planning
Forming and executing action plans for complex tasks.
Category: planning
Multilingual
Competence in many natural languages (from a few to over a hundred): understanding, generation, translation, and code-switching within a single conversation. Frontier models support a wide range of languages with comparable quality.
Category: language

Benchmark results

4 benchmarks
CVE Rediscovery
accuracy
89.6%%
📅 9 Nov 2026📄 anthropic.com/claude/mythos
Mythos 5.2 result per Anthropic product page.
CyScenarioBench
accuracy
76.3%%
📅 9 Nov 2026📄 anthropic.com/claude/mythos
Cybersecurity benchmark.
De novo binder design (24 hr)
accuracy
85.1%%
📅 9 Nov 2026📄 anthropic.com/claude/mythos
De novo binder design (biology).
Terminal-Bench-Science
accuracy
75.3%%
📅 9 Nov 2026📄 anthropic.com/claude/mythos
For comparison Fable 5.2: 74.6%.

Pricing

Technical architecture

Deployment and security

🔒 Security / Enterprise
✓ Verified enterprise information
Updated: 18 Nov 2026↗ Security documentation