Proprietary   Anthropic · Released Returned July 1, 2026 (originally launched March 2026)

Claude Fable 5

The World's Most Capable Model — Returned After Suspension

Architecture Proprietary with Constitutional Reasoning framework
Parameters Undisclosed (frontier proprietary)
Context Window 1 million tokens
Modalities Text, Code, Images, Audio
Input Pricing $10.00 / MTok
Output Pricing $50.00 / MTok
Open Weights No — proprietary, available via Anthropic API, Claude.ai, Claude Code, Claude Cowork
Availability Claude.ai, Anthropic API, Claude Code, Claude Cowork. Mythos 5 restricted to approved US organizations only.
Why This Model Matters Claude Fable 5 is Anthropic's most capable model and currently the strongest publicly available AI system by aggregate benchmark performance. It reclaims the #1 position on the Artificial Analysis Intelligence Index and leads on SWE-bench Pro (80.3%) and SWE-bench Verified (95%). The model's story in July 2026 is unique: on June 12, a US Commerce Department export-control directive required Anthropic to suspend Fable 5 and the even more capable Mythos 5 globally — the first time frontier AI models were taken offline by government order. During the 18-day suspension, Anthropic launched Claude Sonnet 5 (June 30), which became the new default for Free and Pro users. When Fable 5 returned on July 1, it carried a tighter safety classifier and stricter export-use terms. The updated classifier makes the model more likely to refuse borderline requests and more likely to add safety caveats to its responses. Mythos 5, which is even more capable than Fable 5, remains restricted to approved US organizations only. For most professional workflows, Sonnet 5 at $2/$10 per MTok (intro pricing through August 31, then $3/$15) covers 90% of Fable 5's capability at 80% lower cost. Fable 5 at $10/$50 per MTok is the right choice for the hardest 5-10% of tasks where the absolute capability ceiling is necessary.
Visit Anthropic → ← All Releases
Performance

Benchmarks & Evaluation Results

SWE-bench Pro (vendor)

80.3% — highest score

SWE-bench Verified

95% — tied for highest

Terminal-Bench 2.1

Leader position across agentic coding tasks

Artificial Analysis Intelligence Index

#1 overall position

GDPval-AA v2

1,815 — highest score (Fable 5 Max)

In Depth

Full Analysis of Claude Fable 5

Fable 5's 18-day suspension was an unprecedented event in AI history — the first time a government ordered a frontier model taken offline. The June 12 export-control directive from the US Commerce Department applied globally, affecting all users regardless of nationality. Anthropic complied immediately, suspending both Fable 5 and the even more capable Mythos 5. During the suspension, the company launched Claude Sonnet 5 as a temporary replacement on June 30. On July 1, the Commerce Department

lifted the controls, and Fable 5 returned with an updated safety layer featuring a tighter classifier that reduces the model's willingness to engage with borderline requests. Despite these new constraints, Fable 5 reclaims the top position across most benchmark leaderboards. On SWE-bench Pro, it achieves 80.3% — the highest vendor-reported score. On the Artificial Analysis Intelligence Index, it holds the #1 position. The model's 1M-token context window, Constitutional Reasoning framework, an

d enterprise-grade safety features make it the gold standard for organizations that need the absolute best available intelligence and are willing to pay for it. For teams evaluating Fable 5, the key consideration is whether their specific workload requires the capability ceiling that justifies the $10/$50 per MTok price premium, or whether Sonnet 5's $2/$10 intro pricing (which beats Claude Opus 4.8 on Terminal-Bench 2.1) is sufficient.

Assessment

Strengths & Considerations

Strengths

  • + Currently #1 on Artificial Analysis Intelligence Index and SWE-bench Pro
  • + 1M-token context window for processing entire codebases
  • + Constitutional Reasoning framework balances capability with safety
  • + Strongest SWE-bench Verified score at 95%
  • + Enterprise-grade safety and alignment infrastructure
  • + Claude Code and Claude Cowork for developer workflow integration
  • + Most comprehensive safety evaluation and red-teaming of any model

Considerations

  • Most expensive model at $10/$50 per MTok
  • Tighter safety classifier after return may increase refusal rate
  • Mythos 5 (even more capable) restricted to approved US organizations only
  • Export-control risk remains for international deployments
  • Updated export-use terms require review for teams with global operations
  • Sonnet 5 covers 90% of use cases at 80% lower cost
Applications

Best Use Cases for Claude Fable 5

Where this model excels and the types of workloads it is best suited for.

Use Case 1

Most challenging multi-file code migrations and refactoring

Use Case 2

Complex reasoning chains where correctness cannot be compromised

Use Case 3

Highest-stakes agentic operations requiring maximum reliability

Use Case 4

Safety-critical applications needing Constitutional Reasoning framework

Use Case 5

Advanced research analysis requiring frontier intelligence

Use Case 6

Enterprise deployments where capability justifies premium pricing

Getting Started

How to Use Claude Fable 5

Quick Start Guide

Access Fable 5 through Claude.ai, the Anthropic API at api.anthropic.com, Claude Code (terminal-based coding agent), or Claude Cowork (collaborative development tool). Select model ID 'claude-fable-5' for API access. For teams needing even more capability, apply for Mythos 5 access — currently restricted to approved US organizations. For most workloads, start with Claude Sonnet 5 at $2/$10 through August 31.

Visit Anthropic → ← All July 2026 Releases