Introducing Claude Fable 5 and Claude Mythos 5

Introduction

On June 9, 2026, Anthropic announced something it had never done before. Furthermore, for the first time in the company's history, a model from its most capable internal tier Mythos-class became publicly accessible in any form. Claude Fable 5 and Claude Mythos 5 represent the most powerful models Anthropic has ever released, establishing a brand-new capability tier that sits above the Opus class entirely. Consequently, for developers, enterprise teams, researchers, and AI professionals evaluating the current frontier, understanding what these models deliver, how they differ, what they cost, and what they can do is now essential knowledge. This guide covers every dimension of the June 9 launch: capabilities, benchmark performance, pricing details, safety architecture, platform availability, and what the release means for the broader AI landscape.

The Mythos-Class Tier: A New Level Above Opus

Claude Fable 5 and Claude Mythos 5 do not fit within the existing Claude model family structure. They define a new one. Since Claude 3 arrived in March 2024, Anthropic organized its models into three named tiers: Opus at the top for maximum capability, Sonnet in the middle for balance, and Haiku at the base for low-latency, high-volume use. This June 2026 launch introduces a fourth tier Mythos-class that sits above Opus entirely.

The designation reflects a qualitative shift. These are not incremental benchmark improvements over Opus 4.8. According to Anthropic's official benchmark data, Fable 5 surpasses Opus 4.8 by substantial margins across every major evaluation category and the lead grows larger as tasks become longer and more complex.

The Mythos family began in April 2026 with Claude Mythos Preview, a restricted model available only to a handful of vetted cybersecurity partners through Project Glasswing. Its cybersecurity capabilities were advanced enough that Anthropic declined to release it publicly. The June 9 announcement of Claude Fable 5 and Claude Mythos 5 marks the delivery of Anthropic's stated goal: to eventually bring Mythos-level capability to all users, as quickly and as safely as possible.

Claude Fable 5: The Publicly Available Mythos-Class Model

Core Specifications

Claude Fable 5 is Anthropic's most capable widely released model, the first Mythos-class model cleared for general use after Anthropic developed safety classifiers robust enough to prevent misuse in the highest-risk domains.

The model's technical specifications are as follows:

• API Model ID: claude-fable-5

• Context Window: 1,000,000 tokens (1M) by default

• Maximum Output Tokens: 128,000 tokens per request

• Knowledge Cutoff: January 2026

• Launch Date: June 9, 2026

• Availability: Claude API, Claude.ai, Claude Code (web and CLI), Claude Cowork, Claude Platform on AWS, Amazon Bedrock, Google Cloud Vertex AI, Microsoft Azure Foundry

Benchmark Performance

The benchmark gap between Fable 5 and its predecessors is substantial across every measured category:

Software Engineering:

• SWE-bench Verified: 95% (Opus 4.8: 80.8%)

• SWE-bench Pro: 80.3% (Opus 4.8: 53.4%; GPT-5.5: 58.6%)

• FrontierCode Diamond: 29.3%

• Terminal-Bench 2.1: 88.0%

Reasoning and Intelligence:

• GPQA Diamond: 94.1% (Opus 4.8: 91.3%)

• Humanity's Last Exam: 59% (Opus 4.8: 40%)

• GDPval-AA: 1932 (Opus 4.8: 1890)

Other Domains:

• Spatial Reasoning: +38.6% over Opus 4.8

• Tool Use: +17.4% over Opus 4.8

• Legal: +13.3% over Opus 4.8

• Health: 66.0%

• OSWorld-Verified: 83.4%

These benchmarks place Claude Fable 5 and Claude Mythos 5 at or near the top of every major model evaluation in June 2026. Independent rankings note it as number one out of 317 coding models with a composite score of 95 out of 100 on coding-specific leaderboards.

Software Engineering in Production

The most documented production result comes from Stripe, which reported that Fable 5 compressed months of engineering effort into days. Working on a 50-million-line Ruby codebase, the model completed a codebase-wide migration in a single day, a task that would have required a full team working manually for over two months. GitHub's Chief Product Officer noted that it took on complex, long-horizon coding tasks with autonomy and reliability that exceeded previous benchmarks. Cognition's CEO called it the highest-scoring model on FrontierBench with particular strength in long-horizon reasoning and generalization to unfamiliar tools. One vibe-coding benchmark provider reported it as the highest-performing model they had tested, nearly saturating base use cases and building apps in less time with fewer tokens.

Knowledge Work Performance

On Hebbia's Finance Benchmark for senior-level reasoning, Fable 5 achieved the highest score of any model tested, with strong gains in document-based reasoning, chart interpretation, and problem solving. IMC reported the model aced their trading-analysis evaluations across factual lookup, conceptual reasoning, root-cause analysis, and expected-value calculations. One analytics firm reported it as the first model to break 90% on their core benchmark of complex, long-running analytical tasks, a ten-point improvement over Opus 4.8. A physics research firm reported that Fable 5 reached near-GPT-5.5-level results on frontier physics research in 36 hours while using a third of the reasoning tokens.

Vision Capabilities

Fable 5 is the new state-of-the-art model for vision tasks. It extracts precise numbers from detailed scientific figures and can rebuild a complete web application's source code from screenshots alone. Notably, earlier Claude models required elaborate helper harnesses with navigation aids and supplementary tools to play Pokémon FireRed. Fable 5 completed the entire game using only raw game screenshots with no maps, navigation aids, or additional game-state information provided.

Memory and Long-Context Performance

Fable 5 maintains coherent performance across millions of tokens during long-running agentic tasks. It actively improves its own outputs using internally generated notes. When tested on the deck-building game Slay the Spire with persistent file-based memory enabled, Fable 5 benefited three times more from that memory than Opus 4.8 under the same conditions, and reached the game's final act three times more frequently.

Claude Mythos 5: The Restricted Research-Grade Version

Claude Mythos 5 shares the same underlying model, architecture, and training as Fable 5. The single structural difference is that Mythos 5 operates without the safety classifiers that constrain Fable 5's responses in high-risk domains. This gives Mythos 5 the full, unblocked scope of Mythos-class capability including its most advanced cybersecurity skills.

The cybersecurity performance of unblocked Mythos 5 is striking. On ExploitBench a cybersecurity evaluation Mythos 5 scores 78.0%, nearly double Opus 4.8's 40.0%. This capability gap is precisely why Mythos 5 is not generally available. It is offered only to approved customers through Project Glasswing, a program operating in collaboration with the US government that focuses on legitimate defensive cybersecurity and critical infrastructure protection.

Life Sciences Breakthroughs with Mythos 5

In protein design testing, Anthropic's internal experts used Mythos 5 to accelerate aspects of drug design by approximately ten times. With protein design and bioinformatics tools and no human assistance the model matched or exceeded skilled human operators. Working independently, it chose binding sites, selected and ran protein design tools, and recovered from failures. Nine of 14 protein targets in this study yielded strong drug design candidates currently under investigation.

Mythos 5 is also Anthropic's first model to consistently generate novel, compelling scientific hypotheses. In blinded comparisons against Opus-class models, Anthropic's scientists preferred Mythos 5's molecular biology hypotheses approximately 80% of the time. One hypothesis about an E. coli protein mechanism was independently corroborated by a separate research laboratory working on the same problem.

In a separate genomics project, Mythos 5 conducted more than a week of largely autonomous research assembling single-cell data for millions of cells spanning 138 animal species and training a custom machine learning model that outperformed a recently published Science journal model while being 100 times smaller.

Pricing: The Complete Breakdown

Pricing is one of the most precisely documented aspects of this launch. Claude Fable 5 and Claude Mythos 5 are priced identically at every tier.

Standard API Pricing

The US-only inference option applies a 1.1x multiplier on both input and output tokens and is available for teams with data-residency requirements.

Prompt Caching

Prompt caching applies a 90% discount to cached input tokens reducing the effective cost of repeated context significantly for applications with consistent system prompts or large fixed reference documents.

Context Pricing

There is no additional charge for long-context usage. A request using the full one-million-token context window is billed at the same $10 per million input token rate as a short request. This is consistent with Anthropic's stated policy of eliminating long-context surcharges.

Comparison to Other Models

Fable 5 and Mythos 5 cost exactly twice the price of Opus 4.8 and less than half the price of the original Claude Mythos Preview. They are the most expensive models Anthropic currently offers for general use, reflecting their position as the new capability frontier.

Subscription Plan Access and the June 22 Deadline

For users on Claude's subscription plans, Fable 5 operates under a staged availability structure:

June 9-22, 2026: Fable 5 is included on Pro ($20/month), Max ($100/month), Team, and seat-based Enterprise plans at no extra cost. Within these plans, Fable 5 counts as approximately double the usage of Opus meaning it consumes a user's included limits roughly twice as fast.

From June 23, 2026: Fable 5 will be removed from these subscription plans. Using it after this date will require purchasing usage credits until Anthropic has sufficient infrastructure capacity to restore it as a standard included model.

Future restoration: Anthropic has committed to restoring Fable 5 as a standard subscription plan model as quickly as capacity allows, with advance notice before any changes take effect.

For API users and consumption-based Enterprise plans, Fable 5 is fully available from June 9, 2026, at the standard $10/$50 per million token rate with no staged rollout restriction.

Safety Architecture: How Fable 5's Classifiers Work

The Classification System

The architecture that distinguishes Claude Fable 5 and Claude Mythos 5 is a set of separate AI systems classifiers that run alongside the main model in real time. These classifiers detect potential misuse, including jailbreak attempts, before the main model generates a response. When a classifier determines that a query falls into a covered risk category, the request is automatically routed to Claude Opus 4.8 instead. Users receive a notification whenever this fallback occurs.

This design means that users always receive a substantive response, a capable Opus 4.8 answer rather than a flat refusal. Anthropic's early deployment data shows that more than 95% of Fable 5 sessions involve no fallback at all. In those sessions, Fable 5 performs effectively identically to Mythos 5.

The Three Covered Risk Categories

1. Cybersecurity: Mythos-class models demonstrate an exceptional ability to discover and exploit software vulnerabilities and conduct multi-stage agentic hacking including reconnaissance, lateral movement, and exploitation. Classifiers in this category cover both specific exploit requests and offensive cyber tasks in a broader sense.

2. Biology and Chemistry: Advanced capabilities in protein design, molecular biology, and chemical synthesis create dual-use risks. These classifiers block queries where the information could provide meaningful uplift to anyone pursuing biological or chemical harm.

3. Distillation: Technical processes involving the concentration or extraction of dangerous materials are covered by a separate classifier category.

Alignment Assessment

In Anthropic's automated alignment assessment, Mythos 5's level of misaligned behavior including deception and cooperation with misuse was found to be low and similar to that of Opus 4.8. Since both share the same underlying model, Fable 5's alignment profile is comparable.

Platform Availability

Claude Fable 5 is generally available from June 9, 2026, across the following platforms:

• Claude API - directly from Anthropic

• Claude.ai - the consumer and professional chat interface

• Claude Code - web and CLI interfaces

• Claude Cowork - the desktop task automation tool

• Claude Platform on AWS - North America, South America, Europe, Asia Pacific

• Amazon Bedrock - US East (N. Virginia) and Europe (Stockholm) at launch, with additional regions to follow

• Google Cloud Vertex AI

• Microsoft Azure Foundry

Claude Mythos 5 is not generally available. Access is restricted to approved Project Glasswing partners. Organizations seeking access should contact their Anthropic, AWS, or Google Cloud account team.

Both models are designated Covered Models, meaning they carry mandatory 30-day data retention requirements and are not available under zero data retention agreements.

What These Models Mean for AI Professionals

The launch of Claude Fable 5 and Claude Mythos 5 raises the practical capability ceiling for AI in production environments. For professionals across every technical and commercial discipline, building structured expertise in the tools and systems that sit at this frontier creates a durable professional advantage.

For professionals who want to, work effectively with Claude Fable 5 and Claude Mythos 5 understanding how to structure long-context prompts for the one-million-token window, how the classifier fallback architecture affects production behavior, and how to integrate these models into agentic workflows with appropriate safety considerations requires platform-specific depth. Those who pursue a Claude AI Expert certification develop exactly that applied technical knowledge, enabling them to build, evaluate, and deploy Claude Mythos-class applications with genuine architectural confidence.

Additionally, as Claude Fable 5 and Claude Mythos 5 introduce capabilities that now require safety classifiers to deploy responsibly including frontier cybersecurity skills and autonomous scientific research understanding the foundational principles of AI safety, alignment assessment, and responsible deployment is increasingly expected of any technical professional working with AI at a production level. An AI Certification provides the structural technical grounding to evaluate systems like Fable 5 critically, assess their architecture, and advise organizations on responsible adoption at scale.

Furthermore, who want to build recognized expertise across the full technology stack that supports deploying and managing Mythos-class models in production including cloud infrastructure, API integration patterns, model evaluation, and enterprise AI governance a Tech Certification provides verified credentials across the domains that matter most to employers and clients operating at the current AI frontier.

Finally, for marketing professionals, growth strategists, and commercial teams leveraging Fable 5's exceptional knowledge work, document reasoning, and long-horizon analytical capabilities, a Marketing Certification connects these AI capabilities to strategic marketing outcomes providing the framework to translate Fable 5's performance advantages into campaign performance, content operations, and measurable commercial results.

20 Frequently Asked Questions (FAQs)

What are Claude Fable 5 and Claude Mythos 5?

Claude Fable 5 and Claude Mythos 5 are Anthropic's most powerful AI models, launched June 9, 2026. They are the first generation of Mythos-class models, a new tier above the Opus class in capability. Fable 5 is broadly available with safety classifiers. Mythos 5 is the same model without classifiers, available only through Project Glasswing.

When were Claude Fable 5 and Claude Mythos 5 released?

Both models launched on June 9, 2026. Claude Mythos Preview the restricted predecessor had been available since April 2026 through Project Glasswing to a small group of vetted cybersecurity partners.

What is the API model ID for Claude Fable 5?

The official API model ID is claude-fable-5. For Mythos 5, the model ID is claude-mythos-5.

How much does Claude Fable 5 cost?

Standard pricing is $10 per million input tokens and $50 per million output tokens. Batch processing costs $5 per million input tokens and $25 per million output tokens. US-only inference is available at a 1.1x multiplier. Prompt caching applies a 90% discount to cached input tokens.

How does Fable 5 pricing compare to Opus 4.8?

Fable 5 costs exactly twice the price of Opus 4.8 which is priced at $5 per million input tokens and $25 per million output tokens. Both are significantly cheaper than Claude Mythos Preview, which was priced at approximately $25 per million input tokens and $125 per million output tokens.

Is Claude Fable 5 free on subscription plans?

Yes, through June 22, 2026, Fable 5 is included at no extra cost on Pro, Max, Team, and seat-based Enterprise plans. From June 23 onward, usage credits are required until Anthropic restores it as a standard subscription feature when capacity allows.

How does Fable 5 count against subscription usage limits?

Within subscription plans, Fable 5 counts as approximately double the usage of Opus 4.8. A Fable 5 session therefore consumes a user's included monthly limits roughly twice as fast as an equivalent Opus 4.8 session.

What context window do these models support?

Both Claude Fable 5 and Claude Mythos 5 support a one-million-token context window by default. The maximum output per request is 128,000 tokens. There is no additional charge for long-context usage.

What is the knowledge cutoff for Fable 5 and Mythos 5?

The knowledge cutoff date for both models is January 2026.

What is Claude Fable 5's score on SWE-bench?

Fable 5 achieves 95% on SWE-bench Verified, the highest score recorded for this benchmark and 80.3% on SWE-bench Pro, compared to 58.6% for GPT-5.5 on the same evaluation.

How does the safety classifier fallback work?

Separate AI classifier systems run alongside Fable 5 in real time. When a query matches a covered risk category, the request is automatically routed to Claude Opus 4.8 instead. Users are notified whenever a fallback occurs. More than 95% of sessions involve no fallback at all.

What are the three classifier risk categories?

The three covered categories are cybersecurity (discovery, exploitation, and offensive cyber tasks), biology and chemistry (queries that could provide uplift for biological or chemical harm), and distillation (processes involving concentration of dangerous materials).

What is Claude Mythos 5?

Claude Mythos 5 is the same underlying model as Fable 5, without the safety classifiers in covered domains. It has the strongest cybersecurity capabilities of any model in the world according to Anthropic. It is not generally available, access requires approval through Project Glasswing.

Who can access Claude Mythos 5?

Access to Mythos 5 is restricted to approved organizations through Project Glasswing, a program operating in collaboration with the US government. Organizations seeking access should contact their Anthropic, AWS, or Google Cloud account team.

What is the 30-day data retention requirement?

Both Fable 5 and Mythos 5 are Covered Models, meaning all conversation data is retained by Anthropic for 30 days for safety monitoring purposes. Neither model is available under zero data retention agreements.

What platforms support Claude Fable 5?

Fable 5 is available on the Claude API, Claude.ai, Claude Code, Claude Cowork, Claude Platform on AWS, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Azure Foundry.

What was Fable 5's most notable production result?

Stripe reported that Fable 5 completed a codebase-wide migration in a single day on a 50-million-line Ruby codebase, a task that would have taken a full human team over two months.

How does Mythos 5 perform in life sciences research?

Mythos 5 accelerated protein design by approximately ten times in controlled testing, matched or exceeded skilled human operators with no human assistance, and trained a genomics model that outperformed a recently published Science journal model while being 100 times smaller.

What are the batch pricing rates for Fable 5?

Batch processing available for asynchronous workloads with up to 24-hour turnaround is priced at $5 per million input tokens and $25 per million output tokens. This represents a 50% discount over standard real-time API pricing.

What happened to Claude Mythos Preview?

Claude Mythos Preview was the first Mythos-class model, available from April 2026 exclusively to Project Glasswing partners. Claude Mythos 5 supersedes it as an upgrade for existing approved partners. Fable 5 delivers the same Mythos-class capability to the general public for the first time, at less than half the price of Mythos Preview.