We’re rolling out Fable 5 across Pro, Team, and Enterprise tiers, with a free access window from early June to mid-June 2026; after the window, usage caps kick in and it’s best to route heavy tasks to Opus or Sonnet while reserving Fable 5 for agentic work, all within a compliance-conscious framework due to mixed-model responses in logs.

Migration guidance is simple: a one-line code change moves API calls from Opus 4.8 to Fable 5, followed by a practical checklist focused on evaluation, cost controls, and monitoring to prevent drift.

Intro context: Claude Fable 5 debuts under Anthropic’s Mythos tier, outperforming Opus 4.8 in benchmarks and pricing, prompting developers to weigh value against cost.

Product lineup shift: The Mythos tier brings Fable 5 and Mythos 5 in June 2026, repositioning Opus as a mid-tier option and expanding Sonnet 5 as the default free/Pro tier by late June 2026.

Real-world use cases favor Fable 5 for agentic code review, long-document analysis, and regulated environments, while Opus 4.8 remains a cost-efficient choice for high-volume general queries and basic reasoning.

Key specs and pricing: Fable 5 is a Mythos model with a 1,000,000-token context and a 1,000,000-token output window, priced at $10 input and $50 per 1M tokens output, exactly double Opus 4.8, with safety routing directing sensitive prompts to Opus 4.8 and Mythos 5 access limited to vetted customers.

Memory and compaction: Fable 5 uses adaptive thinking and automatic session compaction to sustain longer agent sessions without manual pruning, with cost implications from increased output reasoning.

Safety architecture: Fable 5 automatically routes sensitive prompts to Opus 4.8 while Mythos 5 remains restricted to vetted customers, a modular design praised for regulated environments and auditing.

Benchmark signals: Independent tests show Fable 5 leading Opus 4.8 in coding and agentic tasks, with strong tool-use scores and competitive PhD-level reasoning metrics, reinforcing its edge in multi-step reasoning scenarios.

Best-use guidance: Prioritize Fable 5 for autonomous coding and long-document tasks, reserve Opus 4.8 for general production chat, and use Sonnet 5 when cost sensitivity is paramount, with staged migration for long documents and regulated workflows.