oa-prompting-model-tiering

To choose the right Claude model, match task complexity to model tier: use Claude Haiku for simple tasks, Sonnet for general coding, and Opus for high-complexity requirements. This model selection guidance optimizes cost, speed, and quality for your AI agents.

The best way to specify Claude models in oa CLI operations is using explicit model ID formats like `--model claude/haiku` or `claude/sonnet`. Correct command-line usage patterns prevent common model selection mistakes.

Use Claude Opus instead of Sonnet or Haiku when your AI agent tasks have high complexity and performance requirements. Use the model tiering decision tree to evaluate cost versus quality for optimal model selection.

Yes, you can optimize AI agent costs by model tiering: assign Claude Haiku for simple tasks to save cost, Sonnet for balanced general coding, and Opus only for complex performance needs. This tiering strategy maximizes cost optimization.

Common anti-patterns when selecting Claude models include using incorrect model ID formats, failing to explicitly specify models in oa CLI commands, and defaulting to Opus for simple tasks, which breaks cost optimization and model tiering rules.