Frequently asked questions

Claude Fable 5 is Anthropic's most capable generally available AI model, released on June 9, 2026. Anthropic describes it as "a Mythos-class model made safe for general use" - the first time the company's top model class has shipped to the public. It has a 1M-token context window, 128K output tokens, and set the state of the art on most launch benchmarks, including SWE-Bench Pro at 80.3%. The API model ID is claude-fable-5.

On the Claude API: $10 per million input tokens, $50 per million output tokens - exactly 2x Claude Opus 4.8. The Batch API halves that to $5/$25, and prompt-cache hits cost $1/MTok. In claude.ai paid plans, Fable 5 counts roughly 2x the usage of Opus toward limits. See the full breakdown in our pricing guide.

Temporarily, for subscribers: Fable 5 is included at no extra cost in Pro, Max, Team, and Enterprise plans from June 9 to June 22, 2026. From June 23 it requires usage credits billed at API rates. Anthropic says it aims to restore Fable 5 as a standard part of subscriptions when capacity allows. It is not available on the free plan.

Type /model fable in Claude Code. Fable 5 is not the default model on any plan, and it consumes about 2x the usage of Opus. Our Claude Code guide covers best practices for long-horizon runs.

Use model ID claude-fable-5. No thinking parameter is needed - adaptive thinking is always on. Control depth with output_config: {"effort": "high"}. Note the breaking changes: thinking: {"type": "disabled"}, budget_tokens, non-default sampling parameters, and assistant prefills all return 400 errors. Full details in the API quickstart and migration guide.

The same underlying model as Fable 5, with cyber safeguards lifted - Anthropic calls it "the strongest cybersecurity capabilities of any model in the world." It is restricted to approved Project Glasswing partners: cyberdefenders, critical-infrastructure operators, and government. Read the full Mythos 5 explainer.

A 1-million-token context window at standard pricing (no long-context premium), 128K max output tokens, and a reliable knowledge cutoff of January 2026. More in the context & memory deep dive.

All of them, since day one: Claude API, Claude Platform on AWS, Amazon Bedrock (anthropic.claude-fable-5), Google Vertex AI, Microsoft Foundry, GitHub Copilot, claude.ai paid plans, and Claude Code. Details in our platform availability coverage.

Fable 5 runs real-time safety classifiers for cyber, bio, and reasoning-extraction risks. When one triggers, the API returns stop_reason: "refusal"; with the opt-in fallbacks parameter (or claude.ai's "switch models when a message is flagged" toggle), the request automatically retries on Opus 4.8. This happens in under 5% of sessions. Anthropic concedes the classifiers are "stricter than would be ideal." See how the safety system works.

On launch benchmarks Fable 5 leads on agentic coding (SWE-Bench Pro 80.3% vs 58.6%/54.2%), knowledge work (GDPval-AA Elo 1932), computer use (OSWorld-Verified 85.0%), legal, and health. GPT-5.5 keeps the lead on abstract reasoning (ARC-AGI-2 85.0%, GPQA Diamond 94.4%). Full numbers in the benchmark analysis.

No. Fable 5 is designated a "Covered Model": traffic is subject to mandatory 30-day retention used only for safety review (with logged human access), and it is not available under zero-data-retention agreements. Workloads requiring ZDR should stay on Opus 4.8.

Anthropic hasn't given an official etymology, but the naming follows the new Mythos tier: Mythos for the unrestricted frontier class, Fable for the version whose story can be safely told to everyone. It continues Anthropic's literary naming alongside Opus, Sonnet, and Haiku.