Claude Fable 5.1
Claude Fable 5.1 is Anthropic's model for coding, knowledge work and agentic scientific research, with effort levels, root-cause fixes and EFS privacy.
Model Overview
Capabilities, design details, and architectural traits
Claude Fable 5.1 - Anthropic's model for coding, knowledge work and long-running problem-solving
Claude Fable 5.1 is Anthropic's model for coding and knowledge work, with research capabilities aimed at agentic scientific progress. Its defining trait is that it is the same underlying model as Claude Mythos 5.1, released with different levels of safeguards: Fable 5.1 is generally available, while Mythos 5.1 is restricted to trusted access programs for cybersecurity and life sciences work.
| Differentiator | Detail |
|---|---|
| Twin of Mythos 5.1 | Identical model to Claude Mythos 5.1, differing only in safeguard level; Mythos is limited to trusted access programs |
| Effort levels | Runs at Low, Medium, High, Extra High or Max effort; defaults to High in Claude Code and Medium in Claude Cowork and Claude.ai |
| Root-cause fixes | Avoids easy-seeming shortcuts and fixes root causes rather than symptoms; found a rare crash in Millennium's internal systems that engineers and other models could not explain |
| Vulnerability discovery | Can be used to discover software vulnerabilities, but not to develop exploits for them |
| Enterprise Frontier Safeguards | EFS stores data in customer-controlled cloud infrastructure, giving zero-data-retention-equivalent privacy while preventing adversarial use |
| Safeguard rerouting | Flagged cybersecurity and biology queries are automatically routed to less capable models |
| Self-checking coding | Writes its own tests to verify work and uses vision to check outputs against the original design or goal |
Built for long-running, unattended work
Fable 5.1 targets jobs that take hours or days: multi-day autonomous coding sessions, codebase-wide features, and agents that plan work, recover from failed steps, and keep users updated without being asked. It operates through Cowork, Slack via Claude Tag, browser use, and unattended managed agents on the Claude Platform.
Document and vision work
The model understands diagrams, charts and tables nested in files and PDFs, supporting document-heavy work in finance, legal, analytics and architecture, and applies the same vision capability to evaluate its own coding output.
Benchmark Performance
Independent evaluations · Artificial Analysis
Accuracy & Capability Details
Compare Models Side-by-Side
Evaluate specifications, pricing, and independent benchmark indices
| Model Details | |||
|---|---|---|---|
| General Info | |||
| Provider | Anthropic | Anthropic | Meta |
| Release Date | September 1, 2026 | July 24, 2026 | September 2, 2026 |
| Knowledge Cutoff | - | May 2026 | - |
| Context & Limits | |||
| Context Window | - | 1M | 1M |
| Pricing (per 1M tokens) | |||
| Input Pricing | $10 | $5 | Free Best Input Pricing |
| Output Pricing | $50 | $25 | Free Best Output Pricing |
| Modalities | |||
| Inputs | textimagefile | textimage | textimagefilevideo |
| Outputs | text | text | text |
| Benchmarks (0-100) | |||
| Intelligence Index | 65.7 Best Intelligence Index | 63.1 | 62.1 |
| Coding Index | 81.6 Best Coding Index | 78.0 | 76.3 |
| Agentic Index | 61.3 Best Agentic Index | 59.2 | 59.3 |
GPQA Benchmark
Graduate-level reasoning and expert Q&A evaluation.
Humanity's Last Exam
Extremely difficult logical reasoning and knowledge.
Long Context Reasoning
Logical reasoning over long context windows.
Independent evaluation data provided by Artificial Analysis. To view the latest benchmarks and full details, visit their official site.