Claude Fable 5
Anthropic · Claude 5
Anthropic's most capable widely released Claude model for long-running agents and high-stakes software work.
Overview
Freshness note: Model capabilities, limits, pricing, and availability can change quickly. This profile is a point-in-time snapshot last verified on July 6, 2026.
Claude Fable 5 is Anthropic’s higher-ceiling Claude model for long-running agents, hard reasoning, and demanding software work. Anthropic released Fable 5 on June 9, 2026, suspended access after a reported safeguard bypass, then restored it globally on July 1 after adding new safeguard updates.
The model matters because it is the broadly released member of the Fable/Mythos pair. Mythos 5 has fewer safeguards and is limited to approved Project Glasswing defensive-cybersecurity customers, while Fable 5 is the general-access route for teams that need more capability than Sonnet or Opus but still need a normal commercial deployment surface.
Capabilities
Fable 5 is aimed at work where the model has to stay coherent over long horizons: multi-step software engineering, large codebase analysis, agent workflows with tool calls, complex document review, and difficult enterprise reasoning tasks. Anthropic describes it as the most capable widely released Claude model, with adaptive thinking always on.
The model is also explicitly shaped by safety constraints. The redeployment post says Fable 5 has stronger safeguards than Mythos 5 and that Anthropic now routes some blocked Fable requests to Opus 4.8 instead. For teams, that means Fable is the higher-capability route, but not necessarily the least-restricted route for every cybersecurity workflow.
Technical Details
Current official anchors:
- API ID:
claude-fable-5 - 1M-token context window
- 128K max output
- text and image input, text output
- adaptive thinking always on
- available through Claude API, Claude Platform on AWS, Amazon Bedrock, Google Cloud, and Microsoft Foundry
Anthropic says Fable 5 and Mythos 5 share the same underlying model. The practical difference is release posture and safeguards: Fable is widely released with stronger safeguards, while Mythos is restricted to defensive cybersecurity partners.
Pricing & Access
Base API pricing:
- Input: $10 per 1M tokens
- Output: $50 per 1M tokens
- 5-minute cache writes: $12.50 per 1M tokens
- 1-hour cache writes: $20 per 1M tokens
- Cache hits and refreshes: $1 per 1M tokens
- Batch API input: $5 per 1M tokens
- Batch API output: $25 per 1M tokens
Anthropic’s July 1 redeployment note says Fable 5 returned globally on Claude Platform, Claude.ai, Claude Code, and Claude Cowork. Exact plan, region, and cloud-provider availability can still vary, so teams should verify the active route before migrating production traffic.
Best Use Cases
Choose Fable 5 when Sonnet 5 is not enough and the work justifies premium pricing: long-running coding agents, complex modernization projects, architecture review, high-stakes document reasoning, and enterprise workflows where better follow-through is worth the cost.
Do not use it as a casual default. Sonnet 5 is the better starting point for most workloads, and Opus 4.8 may be a simpler premium fallback when Fable’s safeguards or commercial access create friction.
Comparisons
- Claude Sonnet 5 (Anthropic): Cheaper balanced default for most Claude work; Fable is the higher-ceiling and higher-cost route.
- Claude Opus 4.8 (Anthropic): Public premium model with mature routing; Fable is newer and positioned as more capable for long-running agents.
- Claude Mythos 5 (Anthropic): Shares Fable specs and pricing but has limited Project Glasswing access for defensive cybersecurity.
- GPT-5.5 Pro (OpenAI): Comparable premium OpenAI route when a team wants the OpenAI ecosystem rather than Claude.