AI models can learn speech and text 4x faster using the combined training method.Interleaved speech-text language models show improved learning efficiency.Speech-text interleaving reduces computational cost by up to 4x.Models demonstrate transfer learning between speech and text domains.