Our new Granite 4 Speech model just dropped, and it has taken the #1 spot on the Open ASR (Automated Speech Recognition) Leaderboard! Even more exciting, at 1B parameters, it is our fastest and lightest Granite Speech to date, a fraction of the size of competitors. Granite 4 1B Speech: 🏆 Best in class word error rates on the Open ASR Leaderboard. ⚡️ Lightning fast 1B parameter form factor built on a Granite 4 foundation 💬 Support for English, French, German, Spanish, Portuguese, and Japanese 💪 Built for business. Open, commercial-friendly Apache 2.0 license, Industry-leading transparency. 🔗 Link below in the replies:
The combination of speed and accuracy is exactly what enterprises need for real-world deployment. I'm curious about the inference latency compared to larger models - have you published any benchmarks on actual hardware configurations?
super interesting! any plans to add more languages??
David Cox is there a voiceprinting feature for multispeaker environments?
👏
Great job! Voice is the future UI
Fantastic!! Congratulations David Cox Luis Lastras and team!
Blog: https://huggingface.co/blog/ibm-granite/granite-4-speech Hugging Face: https://huggingface.co/ibm-granite/granite-4.0-1b-speech