IBM has launched Granite Speech 5.0 2B, a new 2-billion parameter open-source speech-language model designed for multilingual transcription and translation. This model achieves the second-highest ranking on the Open ASR Leaderboard with an average English Word Error Rate (WER) of 5.68%. IBM has also updated its Turbo CTC and 4.1 2B Plus models, enhancing performance for edge devices and speaker diarization, respectively. The Granite Speech 5.0 2B model is available for deployment on various platforms including GPUs, Apple Silicon, and CPUs. AI
IMPACT Sets new SOTA on multilingual transcription and translation benchmarks, pressuring competitors like Qwen3-ASR-1.7B.
RANK_REASON Frontier-lab model release with system card.
Read on Mastodon — mastodon.social →
AI-generated summary · Google Gemini · from 3 sources. How we write summaries →