Anthropic has released Claude Fable 5.1, its latest and most capable generally available model for coding, knowledge work and long-running professional tasks. The company also introduced Claude Mythos 5.1, a more restricted version of the same underlying model designed for advanced cybersecurity and biology research.
The main change is not simply better answers. Fable 5.1 is designed to handle complex projects that can take hours, span multiple applications and require the model to plan, use tools, recover from failures and keep working with limited supervision. Anthropic says the model can operate in Cowork, work through requests in Slack, browse the web and run as a managed agent on its platform.
From answering questions to completing projects
Anthropic is positioning Fable 5.1 for tasks that previously required a person to continuously guide an AI system. The model can break down a larger assignment, choose the tools it needs, execute multiple steps and provide updates as it works.
For example, instead of asking Claude to summarize several research documents, a manager could give it a broader assignment: review a set of market reports, compare them with an existing strategy, identify gaps, build a business case and prepare a draft presentation for review.
The same approach applies to coding. Fable 5.1 can work across an entire codebase, conduct code reviews, improve performance, write tests and use vision to compare its output with the original design.
Built for large amounts of context
One of the model’s strengths is its ability to work with document-heavy tasks. Fable 5.1 can understand diagrams, charts and tables embedded in files and PDFs, making it particularly relevant for areas such as finance, legal work, analytics and architecture.
That changes how knowledge workers can use AI. Instead of feeding a model one document or question at a time, teams can increasingly treat it as a research and analysis layer across large collections of information.
For managers, this could mean delegating tasks such as:
- comparing several versions of a strategy or contract;
- reviewing large research packages;
- preparing market or competitor analysis;
- turning internal data into a first business case;
- reviewing a backlog of documents or requests;
- preparing reports and presentations for human approval.
The goal is not necessarily to remove the manager from the process. It is to move the manager higher up the chain — from manually processing information to reviewing the conclusions, challenging assumptions and making the final decision.
Anthropic is also cutting the cost of agentic work
Fable 5.1 is priced at $10 per million input tokens and $50 per million output tokens. More importantly for long-running workflows, cached inputs now cost $0.25 per million tokens — 75% less than with Fable 5.
Anthropic estimates that this can reduce the cost of typical workloads by around 25% and highly agentic workloads by up to approximately 45%.
This matters because AI agents can repeatedly access the same context while working through a task. Lower cache costs make it more practical to give an agent a large body of company information and let it work through a multi-step assignment rather than constantly starting from scratch.
Fable and Mythos share the same underlying model
Claude Mythos 5.1 is based on the same underlying model but is available only to a small group of vetted organizations working in cybersecurity and life sciences. Anthropic says Mythos has stronger capabilities in areas such as cybersecurity and biology, which also create greater risks of misuse.
Fable 5.1 therefore comes with additional safeguards. For example, it can identify software vulnerabilities in source code, but more advanced penetration testing and exploit-generation requests remain restricted. Anthropic has also refined its biology safeguards to reduce false positives while maintaining restrictions around potentially dangerous research.
What this means for managers
The most important shift is that AI is becoming less about generating individual pieces of content and more about taking responsibility for a complete piece of knowledge work.
A useful starting point for companies is to look for workflows that are:
- repetitive but require several steps;
- spread across multiple documents or applications;
- time-consuming but relatively low-risk;
- easy for a human to review before the final decision.
That could include weekly reporting, research, document comparison, CRM or project backlogs, internal analysis and preparation of presentations.
The challenge then becomes less about finding the perfect prompt and more about designing the right AI workflow: what the agent can access, what it can change, where human approval is required and how its work is monitored.
Claude Fable 5.1 is another sign that the AI market is moving in that direction — from assistants that help people complete individual tasks to agents that can take on entire projects and return with the work largely finished.
