GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
TheStageAI
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Run this model inference with full control and performance in your environment.
Model Details
Model Provider
TheStageAI
Model Tree
Input Modalities
Output Modalities
Supported Functionality
TheWhisper-Large-V3 is a fine-tuned, high-performance variant of OpenAI’s Whisper Large V3 model — optimized by TheStage AI for real-time, low-latency, and low-power speech-to-text (ASR) inference across multiple platforms, including NVIDIA GPUs and Apple Silicon (CoreML).
It provides streaming transcription, word timestamps, and scalable performance for use cases like real-time captioning, meetings, and on-device voice interfaces.
TheWhisper is a fine-tuned Whisper model that can process audio chunks of any size up to 30 seconds. Unlike the original Whisper models, it doesn't require padding audio with silence to reach 30 seconds. We conducted quality benchmarking across different chunk sizes: 10, 15, 20, and 30 seconds. For quality benchmarks, we used the multilingual benchmarks Open ASR Leaderboard.
For comprehensive performance and quality benchmarks see TheWhisper.
| Model | Mean WER |
|---|---|
| openai/whisper-large-v3-turbo | 7.81 |
| openai/whisper-large-v3 | 7.45 |
| thewhisper-large-v3-turbo | 7.88 |
| thewhisper-large-v3 | 7.8 |
| Model | Mean WER |
|---|---|
| openai/whisper-large-v3-turbo | 7.61 |
| openai/whisper-large-v3 | 7.22 |
| thewhisper-large-v3-turbo | 7.45 |
| thewhisper-large-v3 | 7.34 |
| Model | Mean WER |
|---|---|
| openai/whisper-large-v3-turbo | 7.63 |
| openai/whisper-large-v3 | 7.29 |
| thewhisper-large-v3-turbo | 7.47 |
| thewhisper-large-v3 | 7.31 |
| Model | Mean WER |
|---|---|
| openai/whisper-large-v3-turbo | 7.61 |
| openai/whisper-large-v3 | 7.32 |
| thewhisper-large-v3-turbo | 7.45 |
| thewhisper-large-v3 | 7.28 |
Clone the repository
bash
git clone https://github.com/TheStageAI/TheWhisper.gitcd TheWhisperInstall for Apple
bash
pip install .[apple]Install for Nvidia
bash
pip install .[nvidia]Install for Nvidia with TheStage AI optmized engines
bash
pip install .[nvidia]pip install thestage-elastic-models[nvidia] --extra-index-url https://thestage.jfrog.io/artifactory/api/pypi/pypi-thestage-ai-production/simplepip install thestage# additional dependenciespip install flash_attn==2.8.2 --no-build-isolationThen generate access token on TheStage AI Platform in your profile and execute the following command:
bash
thestage config set --api-token <YOUR_API_TOKEN>python
import torchfrom thestage_speechkit.apple import ASRPipeline model = ASRPipeline( model='TheStageAI/thewhisper-large-v3', # optimized model with ANNA model_size='S' chunk_length_s=10, token=hf_token) # inferenceresult = model( "path_to_your_audio.wav", max_batch_size=32, return_timestamps="word") print(result["text"])python
from thestage_speechkit.apple import WhisperStreamingPipelinefrom thestage_speechkit.streaming import MicStream, FileStream, StdoutStream streaming_pipe = WhisperStreaming( model='TheStageAI/thewhisper-large-v3', # Optimized model by ANNA model_size='S', # Window length chunk_length_s=10, platform='apple') # set stride in milisecondsmic_stream = MicStream(step_size_s=0.5)output_stream = StdoutStream() while True: chunk = mic_stream.next_chunk() if chunk: approved_text, assumption = streaming_pipe(chunk) output_stream.rewrite(approved_text, assumption) else: breakpython
import torchfrom thestage_speechkit.nvidia import ASRPipeline model = ASRPipeline( model='TheStageAI/thewhisper-large-v3', # allowed: 10s, 15s, 20s, 30s chunk_length_s=10, # optimized TheStage AI engines device='cuda', token=hf_token) # inferenceresult = model( audio="path_to_your_audio.wav", max_batch_size=32, return_timestamps="segment") print(result["text"])python
import torchfrom thestage_speechkit.nvidia import ASRPipeline model = ASRPipeline( model='TheStageAI/thewhisper-large-v3', # allowed: 10s, 15s, 20s, 30s chunk_length_s=10, # optimized TheStage AI engines mode='S', device='cuda', token=hf_token) # inferenceresult = model( "path_to_your_audio.wav", max_batch_size=32, return_timestamps="segment") print(result["text"])git clone https://github.com/TheStageAI/TheWhisper.gitcd TheWhisperpip install .[nvidia]pip install thestage-elastic-models[nvidia] --extra-index-url https://thestage.jfrog.io/artifactory/api/pypi/pypi-thestage-ai-production/simplepip install thestage# additional dependenciespip install flash_attn==2.8.2 --no-build-isolationthestage config set --api-token <YOUR_API_TOKEN>import torchfrom thestage_speechkit.apple import ASRPipeline model = ASRPipeline( model='TheStageAI/thewhisper-large-v3', # optimized model with ANNA model_size='S' chunk_length_s=10, token=hf_token) # inferenceresult = model( "path_to_your_audio.wav", max_batch_size=32, return_timestamps="word") print(result["text"])from thestage_speechkit.apple import WhisperStreamingPipelinefrom thestage_speechkit.streaming import MicStream, FileStream, StdoutStream streaming_pipe = WhisperStreaming( model='TheStageAI/thewhisper-large-v3', # Optimized model by ANNA model_size='S', # Window length chunk_length_s=10, platform='apple') # set stride in milisecondsmic_stream = MicStream(step_size_s=0.5)output_stream = StdoutStream() while True: chunk = mic_stream.next_chunk() if chunk: approved_text, assumption = streaming_pipe(chunk) output_stream.rewrite(approved_text, assumption) else: breakimport torchfrom thestage_speechkit.nvidia import ASRPipeline model = ASRPipeline( model='TheStageAI/thewhisper-large-v3', # allowed: 10s, 15s, 20s, 30s chunk_length_s=10, # optimized TheStage AI engines device='cuda', token=hf_token) # inferenceresult = model( audio="path_to_your_audio.wav", max_batch_size=32, return_timestamps="segment") print(result["text"])import torchfrom thestage_speechkit.nvidia import ASRPipeline model = ASRPipeline( model='TheStageAI/thewhisper-large-v3', # allowed: 10s, 15s, 20s, 30s chunk_length_s=10, # optimized TheStage AI engines mode='S', device='cuda', token=hf_token) # inferenceresult = model( "path_to_your_audio.wav", max_batch_size=32, return_timestamps="segment") print(result["text"])