IBM Releases Granite Speech 5.0-2B, Updates Turbo CTC and Speaker Attribution Models

The new granite-speech-5.0-2b succeeds granite-speech-4.1-2b at the same parameter count and transcribes English, French, German, Spanish, Portuguese, and Japanese, and translates to and from English in those languages plus English to Italian and Mandarin. IBM trained it on 390,000 hours of public audio plus synthetic data for Japanese ASR, keyword biasing, and translation. The model adds language-ID prompting, keyword lists of up to 2,000 terms, and punctuation and capitalization on request, with capitalization F1 ranging from 89.7 to 99.5 across test sets. It also improves English ASR accuracy and robustness to noise, reverberation, and distant microphones, and strengthens long-form conversational translation. It ranks second among top open-weight ASR models on average English word error rate at 5.06% on the Open ASR Leaderboard's public English short-form test sets, as of October 15, 2026. The model is released under Apache 2.0 with support in Transformers, vLLM, llama.cpp, and mlx-audio for Apple Silicon.

The updated granite-speech-5.0-470m-turboctc keeps its 470 million parameter encoder while adding multilingual transcription for English, Japanese, Spanish, German, French, Italian, and Portuguese, native punctuation and capitalization, and more robust performance in reverberant audio. IBM said compiling a single Conformer block and reusing its graph across the stack cut warmup time, pushing Turbo CTC to 19,870 RTFx on the Open ASR Leaderboard, up from about 12,600 RTFx at launch—roughly 1.6 times faster at the same accuracy level.

The updated granite-speech-4.1-2b-plus remains a drop-in replacement with unchanged architecture, prompt format, and API. IBM said word diarization error rate and speaker-attributed word error rate each drop 7%, average timestamp shift is cut by more than half from 210.86 to 98.54 for all speech, word error rate on long-form test sets falls 24% from 13.53 to 10.22, and errors on biased keywords fall 45% with 20-term keyword lists from 11.79 to 6.48.

Editor's Summary

IBM's Granite Speech 5.0 update adds a new 2B-parameter model for high-accuracy multilingual transcription and translation, while improving Turbo CTC for edge use and 4.1-2B-Plus for speaker-attributed meeting recordings. The models are released under Apache 2.0 and target enterprise speech workloads.