Claude Fable 5

Claude Fable 5 is [Anthropic's most capable widely released AI model](https://www.anthropic.com/claude/fable), built for the most demanding reasoning and long-horizon agentic tasks. It belongs to Anthropic's new Mythos-class model tier, sitting above Claude Opus in capability. ## What Claude Fable 5 Is Good For Claude Fable 5 excels at work that is too long, complex, or multi-staged for other models to sustain. Its most notable strengths include: - **Long-horizon agentic tasks**: The model can work for days at a time in an agent harness like Claude Code or Claude Managed Agents, planning across stages, delegating to sub-agents, and checking its own outputs. - **Ambitious coding projects**: It handles large migrations, complex implementations, multi-day autonomous sessions, writes its own tests, and uses vision to check code outputs against design goals. - **Complex knowledge work**: Teams can hand off large research, analysis, or deliverable projects and review completed work rather than supervising every step. - **Vision-heavy tasks**: Fable 5 can extract precise numbers from scientific figures, understand diagrams and charts nested in PDFs, and even rebuild a web app's source code from screenshots alone — making it well-suited for document-heavy fields like finance, legal, and architecture. - **Scientific research**: The model can posit novel hypotheses and assist with accelerating workflows in life sciences research. For developers building AI-powered applications, Fable 5 is available via API and works well alongside tools like [GitHub Copilot](/tools/github-copilot) for coding workflows or [Perplexity](/tools/perplexity) for research-focused pipelines. Compared to predecessors like [Claude 3 Opus](/tools/claude-3-opus), Fable 5 represents a substantial leap in long-context retention and autonomous task execution. ## Who It Is a Good Fit For Claude Fable 5 is best suited for: - **Software engineers and development teams** tackling large-scale codebases, migrations, or multi-day autonomous development sessions. - **Researchers and analysts** who need thorough, multi-stage knowledge synthesis with minimal supervision. - **Enterprise teams** in finance, legal, and architecture dealing with dense document-heavy workflows. - **API developers** building advanced AI integrations that require high capability on complex, open-ended tasks. Consumer access is available to Pro, Max, Team, and Enterprise plan subscribers on claude.ai. API access is available on the Claude API and major cloud platforms. ## Limitations and Where It Falls Short Claude Fable 5 comes with meaningful trade-offs: - **Safety classifiers restrict some domains**: The model includes automated safety classifiers that can decline or reroute requests related to offensive cybersecurity techniques, biology and life sciences content, and certain frontier AI development tasks. Legitimate work in these areas may also be blocked. - **30-day data retention required**: Using Fable 5 requires 30-day data retention for safety monitoring — zero data retention is not available. - **Higher latency on complex tasks**: Individual requests at higher effort settings can run for many minutes, and autonomous runs can extend for hours, requiring adjusted client timeouts and asynchronous handling. - **Higher cost**: At $10 per million input tokens and $50 per million output tokens, it is priced for high-value workloads rather than everyday or high-volume routine tasks. - **Not for offensive security or biology**: The model explicitly does not support offensive cybersecurity or detailed biology/life sciences requests, with a fallback to a less-capable model when those are triggered.