Anthropic releases Claude Fable 5.1 and Claude Mythos 5.1

Anthropic released Claude Fable 5.1 on September 1, 2026, the successor to Claude Fable 5 for long-running agentic coding, knowledge work, and research. Alongside it, Anthropic shipped an identical model with different safeguards, Claude Mythos 5.1, restricted to vetted organizations in cybersecurity and life sciences through the company's Project Glasswing trusted-access program.

What's new

Both models support a 1M-token context window by default, 128k max output tokens, and always-on adaptive thinking. Pricing holds at $10 / $50 per million tokens, the same headline rate as Fable 5, but prompt cache reads drop to $0.25 per million tokens — 0.025x the base input price, versus 0.1x on other Claude models. Anthropic says that change alone cuts costs by an estimated 25% for typical workloads and up to 45% for heavily agentic usage.

On Terminal-Bench-Science 0.1, Fable 5.1 scores 52.6%, more than double Fable 5's 24.7%. Anthropic also cites a 60% reduction in false positives on cybersecurity tasks, tied to improved alignment behavior around resource access and constraint-following, plus a new Enterprise Frontier Safeguards option that enables zero-data-retention deployments through customer-controlled cloud infrastructure (both models otherwise require 30-day data retention).

Other changes in this release wave: tool_choice types any and tool are unsupported on Fable 5.1 and Mythos 5.1 (auto/none still work); thinking blocks the models produce are only replayable to the same model or newer; per-message effort changes and turn-scoped system messages move into beta; and outputs now carry Anthropic's text watermark plus C2PA Content Credentials on supported image/video files generated through the code execution tool.

Fable 5.1 is available on the Claude API, Claude in Amazon Bedrock, Claude Platform on AWS, Claude on Google Cloud (Vertex AI), and Claude in Microsoft Foundry.

Context

This is the second generation of Anthropic's split-release model: a generally available flagship (Fable) paired with a capability-matched but access-gated sibling (Mythos) reserved for sensitive-domain partners who've been vetted for higher-risk use cases like offensive cybersecurity research. Fable 5 and Mythos 5 established that pattern; 5.1 keeps the same public pricing headline while quietly compressing the cost of the caching-heavy, tool-calling-heavy workloads that dominate agentic use.

Why it matters

For developers, the practical story is the caching discount, not the sticker price — agentic workflows that repeatedly replay long context (multi-step coding agents, research loops) are exactly the workloads that benefit most from a 75%-cheaper cache read, and Anthropic is explicit that this is where the 45% savings shows up. The Terminal-Bench-Science jump also signals Anthropic is pushing harder into scientific/research use cases as a differentiator, not just coding. And the continued existence of a separate, access-gated Mythos tier keeps Anthropic's dual-track approach intact: ship frontier capability broadly, but keep the highest-risk cyber/bio-relevant configurations behind a vetting process rather than a paywall.