Claude Fable 5

Claude Fable 5 is Anthropic's generally available deployment of the same underlying weights as the restricted Claude Mythos 5 model. Anthropic positions the model as a new capability tier above Opus: Mythos 5 is the unsafeguarded trusted-access configuration, while Fable 5 adds production safeguards for public use. The system card reports that Mythos 5 is Anthropic's most capable model to date, with state-of-the-art results across software engineering, long-context agentic tasks, vision, knowledge work, healthcare, cyber, and life-sciences evaluations; Fable 5 matches the underlying model where safeguards do not trigger and behaves closer to Claude Opus 4.8 in safeguarded domains. Fable-specific results include SWE-Bench Verified 95.0%, SWE-Bench Pro 80.0%, Terminal-Bench 2.1 84.3% mean reward with 20.9% of trials hitting a safety fallback, OSWorld-Verified 85.0%, FrontierCode Diamond 29.3%, FrontierCode Main 46.3%, GDPval-AA 1932 Elo, GDP.pdf 29.8%, Blueprint-Bench 2 38.6%, and AutomationBench 17.4%. Fable 5 uses classifiers for cybersecurity, biology and chemistry, and distillation attempts; in Claude client apps these requests fall back to the latest Claude Opus model, while the Messages API blocks by default unless developers implement or opt into fallback. Separate competitive-use safeguards target frontier LLM development work and are estimated by Anthropic to affect about 0.03% of traffic. The model was trained on a proprietary mix of public web data, public and private datasets, and synthetic data. Anthropic prices Fable 5 at $10 per million input tokens and $50 per million output tokens, and exposes it through the Claude API as `claude-fable-5`.

Use the benchmark, Arena and pricing sections above as separate evidence. A missing field means the current data snapshot does not support that claim.