Home/Events/Granite Speech 5.0 Turbo CTC Models Released by IBM

Granite Speech 5.0 Turbo CTC Models Released by IBM

Confirmed
Confidence
90%
Impact: 80%
Updated 1h ago

Consensus Brief

IBM has announced the release of two new models in the Granite Speech family, the granite-speech-5.0-470m-turboctc and granite-speech-5.0-470m-turboctc-nc. These models feature 470 million parameters, achieving over 12,600 RTFx on an NVIDIA H200 GPU, allowing for rapid transcription of speech.

Sourced from
Primary: Hugging Face

What Changed Since Last Update

1h ago

The new models are encoder-only designs that provide significantly faster throughput and improved accuracy compared to previous Granite Speech models.

Claim Ledger

4 claims tracked across sources

Confirmed Fact

The noncommercial model scores an aggregate 4.85% WER and the Apache 2.0 model scores 5.00% WER.

Confirmed Fact

Both models can transcribe more than 3.5 hours of speech in one second using batched inference.

Confirmed Fact

The new models are over 20x faster than previous Granite Speech models.

Confirmed Fact

The granite-speech-5.0-470m-turboctc-nc model carries a CC-BY-NC-SA-4.0 license.

Role-Based Impact Analysis

Source Timeline

1 source corroborating