Granite Speech 5.0: Transcribe, Translate, and Attribute more Accurately and Faster

IBM has released Granite Speech 5.0, an open-source model suite designed to improve the accuracy and speed of speech-to-text transcription, translation, and speaker attribution. The models are trained on over 500,000 hours of speech data and are intended to provide developers with more efficient tools for processing long-form audio. By offering these capabilities as open weights, IBM aims to support applications ranging from real-time meeting transcription to automated multilingual content generation.