IBM releases Granite Speech 5.0 speech model family

IBM updated its Granite Speech lineup with the 2B-parameter granite-speech-5.0-2b and two revised models, covering accurate transcription and translation, fast edge transcription and speaker attribution. Trained on 390,000 hours of public audio and supporting six languages, it ships under Apache 2.0 with a 5.06% average English WER, ranking second among open-weight ASR models.

• Hugging Face releases tokenizers v1 with big speed gains2026-09-21
• IBM Q3 sales up 5% to $17.1B, CEO cites AI tailwind2026-10-21
• Train Microduck to walk in Isaac Lab 3.0 and deploy to real robot2026-10-21
• Hugging Face launches SimReady Central with 2,000+ simulation-ready 3D assets2026-10-21
• XHToken releases Spark-X2.6-4B on-device model2026-10-21