Understanding Amazon Bedrock model lifecycle

This article explains Amazon Bedrock's model lifecycle management, covering three states (Active, Legacy, End-of-Life) and strategies for migrating AI applications as foundation models evolve.

• Models progress through Active, Legacy, and End-of-Life states with clear timelines
• Legacy models receive 6 months' advance notice before End-of-Life date
• Extended access phase allows 3+ additional months of usage after Legacy period
• Pricing may adjust during extended access; existing agreements remain unchanged
• Notifications sent via email, AWS Health Dashboard, console alerts, and API
• Migration planning includes assessment, research, testing, and phased deployment phases
• Update API references, request quota increases, adjust prompts for new models
• Use side-by-side comparison, shadow testing, and A/B testing before full migration
• Models remain available minimum 12 months after launch, Legacy for 6+ months

Effective model lifecycle management requires proactive planning, thorough testing, and phased migration strategies to maintain application continuity while adopting improved foundation models.