Anthropic models permit customers to configure reasoning effort levels, with Claude Code defaulting to a medium effort setting