Anthropic has expanded its Claude model family with two new tiers: Claude Mythos, a restricted preview tier for trusted organizations, and Claude Fable 5, a safety-focused variant designed for sensitive use cases. The company's evolving product strategy reveals a shift toward specialized models tailored to different user needs and regulatory requirements, moving beyond the one-size-fits-all approach that has dominated the AI industry.

What Is Anthropic's New Model Hierarchy?

Anthropic's Claude lineup now spans five publicly available models, each serving distinct purposes. At the foundation sits Claude Haiku 4.5, the fastest and most compact option. Claude Sonnet 5 occupies the middle tier, balancing speed and capability. Claude Opus 5 remains the flagship general-purpose model for complex reasoning tasks. Above Opus sits the newly introduced Mythos tier, which includes Claude Mythos Preview and Claude Mythos 5.

The company also released Claude Fable 5, which shares the same underlying model as Claude Mythos 5 but includes additional safety guardrails for biology, cybersecurity, and large language model research and development. This dual-model approach allows Anthropic to offer cutting-edge capabilities while maintaining strict safety controls for high-risk domains.

How Do These New Models Differ From Claude Opus?

The key distinction lies in capability and access restrictions. Claude Mythos Preview, the first model in the Mythos tier, is not available to the general public. Instead, Anthropic is deploying it to a small number of trusted organizations as part of Project Glasswing, an initiative focused on responsible deployment of advanced AI systems. This selective rollout allows Anthropic to monitor real-world performance and gather feedback before broader release.

Claude Fable 5 represents a different approach to capability management. Rather than restricting access entirely, Anthropic implemented safeguards that route certain queries away from Fable 5 to Claude Opus 5 instead. This routing mechanism protects against misuse in sensitive areas like cybersecurity without completely blocking access to the model. According to Anthropic's documentation, these safeguards trigger in less than 5% of sessions on average, though the company acknowledges they occasionally catch harmless requests.

When Did These Models Become Available?

Claude Fable 5 and Claude Mythos 5 were first released on June 9, 2026. However, their availability was short-lived. On June 12, 2026, just three days after launch, Anthropic suspended access to both models to comply with U.S. Department of Commerce export controls. The suspension lasted until June 30, 2026, when the Department lifted those controls. Anthropic restored access on July 1, 2026.

This rapid suspension and restoration cycle highlights the complex regulatory landscape surrounding advanced AI systems. Export controls can significantly impact product timelines and global availability, forcing companies to navigate both technical and legal considerations simultaneously.

How to Access Anthropic's Full Claude Lineup

• Web and Mobile Interfaces: All publicly available Claude models are accessible through Anthropic's web-based chat interface, mobile app, and desktop applications, allowing users to switch between models mid-conversation based on their needs.
• API Access: Developers can integrate Claude models into their applications using Anthropic's API, with specific model identifiers for each version including claude-opus-5, claude-sonnet-5, claude-haiku-4-5-20251001, and claude-fable-5.
• Specialized Tools: Claude is also available through Claude Code for command-line coding tasks, Claude Cowork for knowledge work automation, Claude in Chrome for web browsing, Claude in Excel for spreadsheet tasks, Claude in PowerPoint for presentation creation, and Claude Tag for Slack-based delegation.

What Does This Mean for Enterprise Users?

The expanded model hierarchy gives enterprises more granular control over which AI capabilities they deploy. Organizations handling sensitive data in cybersecurity, biology, or AI research can now choose Claude Fable 5, which provides advanced capabilities while maintaining safety guardrails tailored to their domain. Alternatively, organizations with unique requirements or higher risk tolerance can apply for access to Claude Mythos Preview through Project Glasswing.

For general-purpose tasks, Claude Opus 5 remains the default recommendation. Its position as the primary public model reflects Anthropic's confidence in its safety profile while maintaining broad capability for complex reasoning, analysis, and creative work. Smaller organizations or cost-conscious users can opt for Claude Sonnet 5 or Claude Haiku 4.5 without sacrificing core functionality.

The tiered approach also signals Anthropic's long-term strategy: as more capable models arrive in the coming months, the company plans to improve its safeguards and reduce false positives in routing mechanisms. This suggests that Claude Fable 5's current 5% false-positive rate will likely decrease as the technology matures, making safety-gated models more practical for everyday use.

Why Does Anthropic Need Multiple Model Tiers?

The proliferation of Claude variants reflects a fundamental challenge in AI development: balancing capability with safety. A single model cannot simultaneously maximize performance in all domains while maintaining strict safety controls for high-risk applications. By creating specialized tiers, Anthropic can offer researchers and enterprises the tools they need while protecting against potential misuse.

The export control suspension also underscores why tiered access matters. By restricting the most advanced models to trusted organizations, Anthropic can comply with government regulations while still advancing AI research and development. This approach allows the company to operate globally while respecting national security concerns.

For everyday users, the expanded lineup means more choice without more complexity. Most people will continue using Claude Opus 5 or Claude Sonnet 5 through their preferred interface. Power users and organizations with specialized needs now have options tailored to their use cases, from ultra-fast responses with Claude Haiku to restricted-access research capabilities through Project Glasswing.