IBM has released Granite Speech 5.0 2B, a 2 billion parameter speech model. This model is designed for transcription and translation across six languages and is available under the Apache 2.0 license. It achieves a strong performance, ranking second on the Open ASR Leaderboard with a 5.68% average English word error rate, and includes keyword biasing capabilities for up to 2,000 terms. The model is versatile, capable of running on various hardware including GPUs, Apple Silicon, and local CPUs. AI
IMPACT Provides a new open-source option for speech transcription and translation, potentially lowering barriers for developers.
RANK_REASON Release of an open-source model with performance benchmarks. [lever_c_demoted from research: ic=1 ai=1.0]
Read on Mastodon — sigmoid.social →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →