Claude Platform Updates: API Keys, New Models, and Tooling Enhancements

Fast mode has been removed for Claude Opus 4.6. Requests previously using fast mode will now run at standard speed and be billed accordingly. Users are advised to migrate to Claude Opus 4.8 to continue using fast mode.

Users can now set expiration dates for API keys and Admin API keys directly within the Claude Console, with options for preset durations or custom settings. Keys with at least a 7-day lifetime will trigger email notifications to the creator before expiration. Existing keys remain unaffected.

Claude Sonnet 5 (claude-sonnet-5) is now available with introductory pricing until August 31, 2026. It features a 1M token context window, 128k max output tokens, and supports the same tools as Sonnet 4.6, excluding the Priority Tier. Key changes include adaptive thinking enabled by default and the removal of manual extended thinking and non-default sampling parameters.

Session event streams for Claude Managed Agents now support event deltas, providing previews of agent messages as they are generated. The API also supports backward pagination for listing sessions and allows overriding agent configurations, such as model or system prompt, on a per-session basis.

Vaults for Claude Managed Agents now include an injection_locationsetting for environment variable credentials, controlling substitution into request headers or body. Webhooks have been expanded to cover agent, deployment, and deployment run lifecycle events, enabling reactive workflows without polling.

Rate limits across the Claude API have been raised, with Sonnet and Haiku limits now matching Opus at all tiers. Usage tiers have been consolidated into 'Start', 'Build', and 'Scale', ensuring no organization receives lower limits.