Use Shorter Model Aliases

If you frequently use long model names in your prompts or code (e.g., 'gpt-3.5-turbo-instruct-0914'), consider using shorter aliases if your framework allows, or use variables to store them. This can slightly reduce token count in complex prompts.