Claude Fable 5.2
Anthropic's Claude Fable 5.2 runs days-long unattended agents, orchestrates 1000+ subagents, verifies its own work, and serves as Advisor or Orchestrator.
Model Overview
Capabilities, design details, and architectural traits
Claude Fable 5.2 - Anthropic's model for daunting technical work
Claude Fable 5.2 is Anthropic's model for the hardest technical problems. It takes on ambiguous, multi-step work, runs unattended for hours to days, and questions and verifies its own output as it goes. Anthropic positions it for ambitious, open-ended projects rather than everyday coding.
| Trait | Detail |
|---|---|
| Same model as Claude Mythos 5.2 | Fable carries strong cybersecurity and biology safeguards so it is open to all users; Mythos 5.2, without those safeguards, is restricted to verified cyber and life-sciences organizations |
| Days-long unattended agents | In an internal test it coordinated over 1,000 subagents for eight days to build a working operating system from an empty folder, complete with drivers, file storage, and an installer |
| Advisor and Orchestrator roles | Haiku, Sonnet, and Opus models consult it on hard problems, and it can delegate simpler tasks back down to smaller models |
| Self-verification loop | It catches its own errors, builds controls on long runs, and flags results that do not reproduce |
| Scientific feats in prerelease | Formalized the proof of Fermat's Last Theorem in 30 hours, solved two open problems in black hole physics, and designed three working electronic devices from scratch |
Where Anthropic says to use it
Anthropic recommends starting with Opus 5.5 for most coding and knowledge tasks, then moving up to Fable 5.2 only when scope and complexity demand it. Its documented use cases span agentic coding across multi-repository monorepos, scientific research, financial analysis, and design for 3D, hardware, and motion.
Fable and Mythos: one model, two access paths
Fable 5.2 and Mythos 5.2 share the same underlying model. The difference is safeguards: Fable's cyber and biology protections make it generally available, while Mythos 5.2 is offered through Anthropic's Cyber and Life Sciences Verification Programs for authorized security engineering, red teaming, and biomolecular design work.
Benchmark Performance
Independent evaluations · Artificial Analysis
Compare Models Side-by-Side
Evaluate specifications, pricing, and independent benchmark indices
| Model Details | |||
|---|---|---|---|
| General Info | |||
| Provider | Anthropic | Anthropic | |
| Release Date | November 9, 2026 | October 28, 2026 | September 22, 2026 |
| Knowledge Cutoff | - | - | - |
| Context & Limits | |||
| Context Window | 1M | 1.0M Best Context Window | 1M |
| Pricing (per 1M tokens) | |||
| Input Pricing | $8 | $2 Best Input Pricing | $4 |
| Output Pricing | $40 | $10 Best Output Pricing | $20 |
| Modalities | |||
| Inputs | textimagefile | text | textimagefile |
| Outputs | text | text | text |
| Benchmarks (0-100) | |||
| Intelligence Index | 64.6 Best Intelligence Index | 60.0 | 59.1 |
| Coding Index | - | - | - |
| Agentic Index | - | - | - |
Humanity's Last Exam
Extremely difficult logical reasoning and knowledge.
Long Context Reasoning
Logical reasoning over long context windows.
SciCode Benchmark
Scientific coding and mathematical modeling.
Independent evaluation data provided by Artificial Analysis. To view the latest benchmarks and full details, visit their official site.
Explore more from Anthropic
Other models by Anthropic
Claude Opus 5.5
Claude Sonnet 5.5
Claude Fable 5
Claude Fable 5.1
Top AI Models
Leading alternatives by intelligence score
