Anthropic launches Claude Fable 5.1 — stronger benchmarks and cheaper long-running agents
Anthropic has released Claude Fable 5.1, its new flagship model for difficult reasoning, coding, research and long-horizon agentic work. It is available now across Claude’s platforms and via the API, including AWS, Google Cloud and Microsoft Azure.
A quick look at Anthropic’s published tests (Fable 5.1 / Fable 5 / Opus 5):
• Terminal-Bench-Science 0.1: 52.6% / 24.7% / 29.0% • Terminal-Bench 4.0: 55.8% / 42.0% / 52.3% • GDPval-AA v2: 1853 / 1723 / 1824 • AutomationBench: 31.4% / 17.1% / 26.9% • CursorBench 3.2: 73.4% / 70.5% / 70.0%
The biggest leap is in agentic scientific research: Fable 5.1 more than doubles Fable 5’s result. It also leads Opus 5 in all five examples above. These are Anthropic’s own evaluations, and the company notes that production safeguards and fallback behavior can affect some scores.
The model keeps Fable 5’s base API price—$10 per million input tokens and $50 per million output tokens—but cached reads are now 75% cheaper at $0.25 per million. Anthropic estimates roughly 25% lower costs for typical workloads and up to 45% for context-heavy agentic tasks.
The trade-off: Opus 5 remains half the base price at $5/$25 and is still Anthropic’s recommended starting point for most workloads. Fable 5.1 is the premium option when maximum reasoning and sustained autonomous work matter more than price.
Source: Anthropic’s announcement and model documentation.