IBM Releases Two Granite Speech 4.1 2B Models: Autoregressive ASR with Translation and Non-Autoregressive Editing for Fast Inference

IBM released Granite Speech 4.1 2B and Granite Speech 4.1 2B-NAR on Hugging Face under Apache 2.0.

The standard model supports multilingual ASR and speech translation, while the NAR version is optimized for faster inference.

IBM also introduced Granite Speech 4.1 2B-Plus for speaker attribution and timestamps.

The open models aim to give enterprises lower-latency speech tools with strong benchmark performance for production use.