Robots Atlas>ROBOTS ATLAS
Claude Mythos 5

Claude Mythos 5

5 · Family: Claude
Anthropic's most capable model for cybersecurity and biology research (SOTA). Available via trusted access programs. Sibling: Claude Fable 5 with safeguards.
✓ Active⏳ Limited accessFeaturedReasoning modelSpecialized AIMultimodal📁 Claude
Context window
Nieujawnione publicznie (rodzina Claude 4/5 zwykle 200K tokenów)
tokens
Parameters
Nieujawnione publicznie
parameters
Release date
9 June 2026
Access:APIDeployment:☁ Cloud

Overview

Claude Mythos 5 is Anthropic's most capable model dedicated to cybersecurity and biology research, released on June 9, 2026. It succeeded Mythos Preview (April 2026), bringing noticeable gains on cybersecurity, biology and healthcare benchmarks. Because of the dual-use risk, access is limited to a small set of vetted partners through trusted access programs.

Mythos 5 shares the same underlying model with Claude Fable 5 — Fable is its public variant with robust safeguards: queries touching high-risk domains (cybersecurity, dual-use biology) are automatically routed to Claude Opus 4.8. Mythos 5 does not apply these filters and exposes full capability in these areas.

Under Project Glasswing, the initiative to secure the world's most critical software (partners: AWS, Anthropic, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, Linux Foundation, Microsoft, NVIDIA, Palo Alto Networks), the earlier Mythos Preview and about 50 partners found more than 10,000 high/critical severity vulnerabilities across systemically important software. A notable individual example: Mozilla found and patched 271 Firefox vulnerabilities using Mythos Preview within two weeks of limited release.

The availability timeline reflects the weight of these capabilities: on June 12, 2026 the US Department of Commerce suspended access to Mythos-class models under a directive concerning foreign nationals (including Anthropic employees). On June 26, 2026 restrictions were lifted for Mythos 5 for over 100 US organizations. On July 1, 2026 the Trump administration lifted the ban entirely.

Pricing: 10 USD per million input tokens and 50 USD per million output tokens. Use requires accepting a 30-day data retention policy for safety monitoring. Anthropic runs extensive safety evaluations — the full system card describes safety test results. Mythos 5's capabilities in cybersecurity and biology are advanced enough that they could be used for both beneficial and harmful purposes, which justifies the trusted-access model.

Classification
Reasoning modelSpecialized AIMultimodal
Family: Claude
Access & deployment
API
Cloud
Weights: Closed
Key parameters
📏 Context: Nieujawnione publicznie (rodzina Claude 4/5 zwykle 200K tokenów)
🧩 Parameters: Nieujawnione publicznie
Tools
📥 Input: text, image, documents

Technical specification

Context window
Nieujawnione publicznie (rodzina Claude 4/5 zwykle 200K tokenów)
tokens
Parameters
Nieujawnione publicznie
parameters
License
Proprietary (Anthropic) — trusted access programs, 30-dniowa retencja danych dla safety monitoring
Hardware requirements
Not applicable — API-only access. Model hosted on Anthropic infrastructure and cloud partners.
Features:Tool use
Modalities
⬇ Input
textimagedocuments
⬆ Output
textcodestructured_data

Capabilities and applications

Native model capabilities
Reasoning
The model's ability to reason logically and solve complex problems.
Category: reasoning
Multi-step reasoning
Carrying out multi-step chains of reasoning across long, complex tasks.
Category: reasoning
Planning
Forming and executing action plans for complex tasks.
Category: planning
Coding
Generating, analysing and modifying source code.
Category: coding
Long context
Maintaining coherence and focus across very long input context.
Category: language
Structured output
Producing data in structured formats such as JSON.
Category: structured_generation

Benchmark results

1 benchmark
CyberSecEval 2
accuracy · cybersecurity — Mythos 5 flagship domain
SOTA%
📅 9 Jun 2026📄 anthropic.com/claude/mythos
Claude Mythos 5 is state-of-the-art in cybersecurity per Anthropic's product page; exact values are disclosed in the system card.

Pricing

Deployment and security

🔒 Security / Enterprise
✓ Verified enterprise information
Updated: 13 Jul 2026↗ Security documentation