Claude Fable 5 and Mythos 5 ship; 1M context and adaptive thinking

We've launched Claude Fable 5 (claude-fable-5), our most capable widely released model, alongside Claude Mythos 5 (claude-mythos-5) for Project Glasswing participants. Both models support a 1M token context window by default, 128k max output tokens, and always-on adaptive thinking.

Key features:

• Use the new tokenizer (introduced with Claude Opus 4.7): same text produces roughly 30% more tokens
• Run safety classifiers on requests and during response generation; returns stop_reason: "refusal" when declined
• Opt-in fallbacks parameter re-runs refused requests on another model
• stop_details.category field includes "reasoning_extraction" for Terms of Service violations
• Adaptive thinking is the only thinking mode; thinking: {"type": "disabled"} not supported
• thinking.display defaults to "omitted" (set to "summarized" for readable thinking summaries)
• Requires 30-day data retention; not available under zero data retention
• Claude Managed Agents now supports scheduled deployments
• Claude Managed Agents vaults support environment variable credentials
• session.thread_* webhook events include session_thread_id field
• Released Swift package in beta for Apple Foundation Models framework