Claude Fable 5: Pricing, Safeguards, Coding Fit, and Limits

Full Guide

Kingy verdict: Claude Fable 5 is a high-cost Anthropic model aimed at long-running coding and knowledge-work projects where review time matters more than minimum token price. Its most consequential buying constraints are the $10/$50 API rate, mandatory safety routing for some sensitive requests, and 30-day safety-monitoring retention—not a provider benchmark leaderboard.

What changed

Anthropic introduced Fable 5 on June 9, 2026 as a generally available Mythos-class model with safeguards intended to make that capability suitable for broad use. The company positions it for multi-stage coding, research, document analysis and asynchronous agent work. Fable and the restricted Mythos 5 share an underlying model; the practical difference is that Fable applies additional safeguards in areas such as cybersecurity and biology.

Availability briefly changed after launch. Anthropic suspended Fable 5 on June 12 following new export controls, then announced on June 30 that those controls had been lifted and global access would resume July 1 across the Claude Platform, Claude.ai, Claude Code and Claude Cowork. That history is relevant because it shows model access can be shaped by policy as well as capacity.

Pricing, access and safeguards

Anthropic lists API pricing at $10 per million input tokens and $50 per million output tokens, with its existing 90% prompt-caching input discount. The API model identifier is claude-fable-5. The product page also states that Fable use requires 30-day data retention for safety monitoring. Buyers with stricter data-handling requirements should confirm how that rule interacts with their plan, region and workload before moving sensitive repositories or documents.

Some cybersecurity and biology queries are routed to Claude Opus 4.8. Anthropic says the safeguards can catch benign work, and API customers need to configure the company’s fallback mechanism. That makes routing behavior part of acceptance testing: record which model handled the request rather than assuming every response came from Fable 5.

How to evaluate it

Use a project that takes several stages and has an auditable finish line: a repository migration, a document-heavy research memo, or an implementation that must pass tests and visual review. Track accepted changes, regressions, retries, total input and output tokens, human review minutes and any safeguard-triggered fallback. Compare the completed artifact at equivalent review standards—not a single impressive response.

Anthropic publishes strong benchmark and customer results, but those are provider-selected evidence. Kingy did not reproduce the evaluations or run a controlled Fable suite. The profile therefore treats benchmark claims as reasons to test, not as an independent ranking.

Primary sources

• Anthropic launch announcement
• Anthropic Fable 5 product page
• Anthropic redeployment update
• Anthropic system-card index
• Claude Platform pricing documentation