Anthropic, a leading AI safety and research company, has just made a significant announcement that's set to reshape the landscape of advanced AI model access. On June 9, 2026, the company officially released two new models: Claude Fable 5 and Claude Mythos 5.
While both models share the same powerful underlying "Mythos-class" architecture, their key differentiator lies in their safety safeguards and intended access. Claude Fable 5 is now generally available with robust classifiers designed for broader public use, making cutting-edge AI capabilities accessible with a focus on safety.
In contrast, Claude Mythos 5, which represents the same core model but with specific cyber safeguards lifted, remains in a limited release program. Its deployment is primarily through Project Glasswing, a collaborative initiative focused on defensive cybersecurity.
Understanding the Mythos-Class Tier
The introduction of the "Mythos-class" tier signifies a new benchmark in AI capabilities, surpassing Anthropic's previous "Opus-class" models. These models are designed for tackling the most ambitious, complex, and long-running tasks that prior AI systems struggled to sustain.
Anthropic states that Fable 5 and Mythos 5 are state-of-the-art across nearly all tested benchmarks of AI capability, excelling in areas such as software engineering, knowledge work, vision, and scientific research.
Claude Fable 5: Advanced AI for General Use
Claude Fable 5 is Anthropic's most capable model released for general availability. It offers exceptional performance in various domains:
- Software Engineering: Fable 5 has demonstrated remarkable prowess in coding tasks. For instance, Stripe reported that Fable 5 compressed months of engineering work into days, performing a codebase-wide migration in a 50-million-line Ruby codebase in a single day—a task that would have otherwise taken a team over two months. It also scores highest among frontier models on Cognition's FrontierCode evaluation for difficult coding tasks.
- Knowledge Work: The model shows strong performance on complex analytical tasks, achieving the highest score on Hebbia's Finance Benchmark for senior-level reasoning. This includes significant gains in document-based reasoning, chart and table interpretation, and problem-solving.
- Vision Capabilities: Fable 5 can understand diagrams, charts, and tables embedded in files and PDFs, making it highly effective for research and document-heavy work in fields like finance, legal, analytics, architecture, and gaming. In coding, it can implement designs with high fidelity and use vision to critique its output.
- Long-running, Asynchronous Execution: Fable 5 can handle complex tasks that previous models could not sustain, operating for days on coding and knowledge work without human intervention. It can plan its approach, check progress against goals, and refine its work autonomously.
The core innovation allowing Fable 5 to be generally available is its advanced safeguard system, which includes classifiers. These classifiers detect queries related to high-risk domains such as cybersecurity, biology, chemistry, and health. When such queries are identified, Fable 5 will instead provide a response from Anthropic's next-most-capable model, Claude Opus 4.8.
Anthropic acknowledges that these safeguards are conservatively tuned and may sometimes catch harmless requests, triggering in less than 5% of sessions on average. This cautious approach allows for the safe and rapid release of such a powerful model to a wider audience.
Claude Mythos 5: Unrestricted Power for Vetted Partners
Claude Mythos 5 shares the identical underlying "Mythos-class" model as Fable 5 but with its cyber safeguards lifted in certain areas. This makes it Anthropic's most potent model for cybersecurity tasks, designed for highly specialized and sensitive applications.
Due to its unrestricted capabilities, Mythos 5 is not publicly available. Instead, it is deployed through a limited, trusted access program, primarily via Project Glasswing.
Project Glasswing: Securing Critical Software with AI
Project Glasswing is a crucial initiative launched by Anthropic in April 2026, in collaboration with the US government and a consortium of leading tech companies, including Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA, and Palo Alto Networks.
The project's goal is to secure the world's most critical software by leveraging advanced AI to find and fix vulnerabilities. The initial model used for this project was Claude Mythos Preview, which has now been upgraded to Claude Mythos 5.
Through Project Glasswing, Mythos-class models have already identified thousands of high-severity vulnerabilities across various operating systems and web browsers. For example, Mythos identified 23,000 critical vulnerabilities, including a 27-year-old flaw in OpenBSD and a 16-year-old bug in FFmpeg. It also found 271 vulnerabilities in Firefox, which is ten times more than Claude Opus 4.6.
This initiative highlights the dual nature of highly capable AI: while it presents potential risks if misused, its power can also be harnessed for profound good, particularly in defensive cybersecurity.
Shared Core, Different Deployment Philosophy
The decision to release two versions of essentially the same powerful model—Fable 5 for general use with safeguards, and Mythos 5 for restricted, specialized use without them—underscores Anthropic's commitment to responsible AI development.
Both Fable 5 and Mythos 5 boast a 1 million token context window, allowing them to process and understand extremely long and complex inputs. They also support up to 128,000 output tokens per request.
The models exhibit enhanced capabilities in agentic autonomy, meaning they can plan, execute, and validate multi-step tasks without constant human intervention. They can self-correct, backtrack when necessary, and explain their reasoning, making them ideal for complex workflows in software development, research, and knowledge work.
Pricing and Accessibility
Anthropic has priced both Claude Fable 5 and Claude Mythos 5 at $10 per million input tokens and $50 per million output tokens. This pricing is less than half the cost of the previous Claude Mythos Preview, making these advanced capabilities more accessible.
For developers, Claude Fable 5 is available via the Claude API (using the model ID claude-fable-5) and through major cloud marketplaces like Amazon Bedrock and Microsoft Foundry.
For users on Claude.ai subscription plans (Pro, Max, Team, and seat-based Enterprise plans), Fable 5 is included at no extra cost from June 9 to June 22, 2026. After this period, it will require usage credits until capacity allows it to return as a standard feature. Anthropic suggests that while Fable 5 is powerful, for shorter or less complex tasks, models like Claude Opus 4.8 or smaller models might be more cost-effective.
The Impact on Tech Professionals
The release of Fable 5 and Mythos 5 marks a significant moment for tech professionals, particularly those in software development, cybersecurity, and scientific research. These models promise to accelerate complex tasks, automate multi-stage workflows, and provide unprecedented analytical capabilities. The ability of Fable 5 to handle long-running, asynchronous tasks means developers can delegate more ambitious projects to AI agents, freeing up human engineers for higher-level strategic work.
For cybersecurity experts, the existence of Mythos 5, even in its restricted form, signals a new era in defensive capabilities. The insights gained from Project Glasswing using Mythos-class models will undoubtedly contribute to building more secure software infrastructure globally.
Anthropic, founded in 2021 by former members of OpenAI including Dario and Daniela Amodei, continues to emphasize AI safety and research, aiming to build reliable, interpretable, and steerable AI systems. The careful rollout of Fable 5 and Mythos 5 exemplifies this core philosophy, balancing cutting-edge capabilities with robust safety considerations.
As AI models become increasingly powerful, Anthropic's approach of providing a generally available, safeguarded version alongside a highly restricted, unrestricted version for critical applications sets an important precedent for the industry. It highlights the ongoing challenge and responsibility of deploying frontier AI while mitigating potential risks.
