Anthropic Launches Claude Fable 5: A Fifth-Generation Model for Extended Autonomous Work

Anthropic has officially unveiled Claude Fable 5, its fifth-generation flagship large language model, engineered to take on the hardest knowledge work and coding challenges with unprecedented autonomy and depth. The model began rolling out on June 9, 2026, and is available to Pro, Max, Team, and Enterprise users, as well as developers through the Claude Platform, cloud marketplaces, and major cloud providers including Amazon Web Services, Google Cloud, and Microsoft Foundry.

Pricing and Availability

Claude Fable 5 is priced at $10 per million input tokens and $50 per million output tokens. The existing 90% input token discount for prompt caching still applies. For workloads requiring US-only inference, 1.1x pricing is an option. The API model identifier is claude-fable-5.

Anthropic reported a brief service disruption on June 12, 2026, when access to Claude Fable 5 became unavailable; access was restored by July 1, 2026.

Capabilities and Use Cases

Claude Fable 5 is described as a Mythos-class model optimized for long-running, asynchronous tasks. Key capabilities include:

• Agents: The model can run in agent harnesses such as Claude Code or Claude Managed Agents for days at a time, planning across stages, delegating to sub-agents, and checking its own work.
• Coding: It supports large migrations, complex implementations, and multi-day autonomous sessions, including self-testing and vision-based output validation.
• Enterprise workflows: Teams can hand off complex, multi-stage knowledge work with minimal oversight, reviewing completed deliverables rather than supervising every step.
• Vision: The model understands diagrams, charts, and tables in files and PDFs, enhancing work in finance, legal, analytics, and architecture.

Safety and Safeguards

Claude Fable 5 includes robust safeguards for cybersecurity and biology. Many queries in these domains are automatically routed to the previous model, Claude Opus 4.8, if flagged. Users are not charged Fable prices for rerouted requests. Anthropic has published an accompanying system card detailing safety evaluations.

Data retention is set to 30 days for safety monitoring purposes.

Benchmark Performance

Anthropic claims Claude Fable 5 achieves state-of-the-art results across coding, knowledge work, vision, and computer use benchmarks. Notable results cited by partners include:

• Cursor: highest score on CursorBench, enabling long-horizon problems previously out of reach.
• GitHub: exceeded previous benchmarks for long-running coding tasks.
• Replit: strongest Claude model results on agentic coding and prototyping.
• Cognition: highest score on FrontierBench for long-horizon reasoning and generalization.
• Scale AI: scored over 90% on a core analytics benchmark, a 10-point jump over Opus 4.8.

Customer Feedback

Several partners provided testimonials. Amplitude noted that Claude Fable 5's reasoning operates at a “senior research scientist grade.” Replit called it a “real step forward” for developers. Other customers, including Hebbia, Block, and Notion, highlighted gains in finance, enterprise analytics, and coding workflows.

Frequently Asked Questions

Anthropic positions Claude Fable 5 as its most capable generally available model, while cautioning that its advanced capabilities in cybersecurity, biology, and chemistry could be misused. The company plans to offer trusted access programs for vetted partners interested in using the model for cybersecurity and biology research. API customers must configure a new Fallback API to handle routed queries.