A 16-bit floating-point format that balances precision and memory efficiency, commonly used for training and deploying large language models.