Tasks and inputs
Granite Speech 5.0 TurboCTC is listed for Automatic speech recognition. It accepts audio and produces text. Use the publisher’s input format and preprocessing for your chosen deployment.
SyncDev model directory · Profile updated Oct 8, 2026
Granite Speech 5.0 TurboCTC is a compact English speech-recognition checkpoint using non-autoregressive CTC decoding.
The specification score summarizes recorded features and limits. It is not a benchmark result.
Recorded modality support shows what this model can accept and produce. An undocumented modality is shown as unknown rather than assumed to be unsupported.
Accepts
Produces
Granite Speech 5.0 TurboCTC is a compact English speech-recognition checkpoint using non-autoregressive CTC decoding.
TurboCTC is a roughly 470M-parameter acoustic model for English transcription. It uses a conformer encoder, Connectionist Temporal Classification and greedy non-autoregressive decoding. The publisher positions it for low-latency or high-throughput speech processing, including local and edge devices.
Use the model’s processor and the sample rate it expects. Batch padding, channel conversion and chunk boundaries can affect recognition, so make these settings consistent across an evaluation. The native Transformers path documented by the publisher requires a version that supports this Granite architecture.
Test a quiet recording, background noise, reverberation and names from the application’s domain. Compare against a reference transcript and inspect substitutions that change meaning. Measure total processing time on the actual device; a GPU benchmark from the model card is not a laptop latency guarantee.
The model uses Apache 2.0 licensing and returns transcription text rather than dialogue or synthesized audio. Speaker labeling is a separate capability and should not be inferred from ASR support. The checkpoint does not define a managed hourly or token rate. Budget device memory, concurrency and any downstream punctuation or diarization work separately from the recognition model.
Compare the model on the work you need to complete.
Granite Speech 5.0 TurboCTC is listed for Automatic speech recognition. It accepts audio and produces text. Use the publisher’s input format and preprocessing for your chosen deployment.
Measure a completed task using the prompts, documents and tools your application will use. Include reasoning and retries in the result.
The weights are downloadable under Apache 2.0. Check runtime support and size the complete model and working context for your expected concurrency.
These records share a directory category. Check the source card, output type and deployment requirements before treating a model as a replacement.
Gemini 3.5 Transcribe Live from Google: Speech transcription model for accurate audio-to-text and captioning workflows.
xAI
Grok Voice STT 1.0 from xAI: Grok Voice STT 1.0 is xAI's speech-to-text model. It supports transcription with word-level timestamps, optional speaker diarization, and multichannel audio.
OpenAI
GPT Realtime Whisper from OpenAI: Streaming speech-to-text model for low-latency transcript deltas from live audio.
openai
openai/whisper-large-v3-turbo is a source-linked Hugging Face repository assigned to Automatic speech recognition. Its recorded task interface accepts audio and produces text.
OpenAI
Whisper 3 Large from OpenAI: Open Whisper checkpoint for robust multilingual transcription and captioning.
OpenAI
Whisper Large v3 Turbo from OpenAI: Speech transcription model for accurate audio-to-text and captioning workflows.