This is the model card of a transformers model that has been pushed on the Hub. Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. Use the code below to get started with the model. Carbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).
Whisper is a pre-trained model for automatic speech recognition (ASR) and speech translation. Trained on 680k hours of labelled data, Whisper models demonstrate a strong ability to generalise to many datasets and domains without the need for fine-tuning.
Parameters242M
Context—
Weights3.9 GB
Licenseapache-2.0
AccessOpen weights
Monthly Downloads2.9M
Runs On
What it takes to serve whisper-small (242M parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them.
| Precision | Weights | Memory needed | Cheapest setup | Per hour | Also fits |
|---|---|---|---|---|---|
| 16-bit | 0.5 GB | 0.6 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 8-bit | 0.2 GB | 0.3 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 4-bit | 0.1 GB | 0.1 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the SAVRN Index, read Sep 18, 2026.
SAVRN's Notes on whisper-small
Six tenths of a gigabyte at 16-bit, 0.1 GB at 4-bit: memory is not the question with whisper-small, OpenAI's 242M-parameter model for speech recognition and speech translation. The cheapest Index listing, one MI300X with 192 GB at $1.85 an hour on-demand, would sit nearly idle, so we run a model this size beside other work or fan many audio streams across one card. Disk is the bigger figure: 16 files totaling 3.87 GB, stored in float32.
Apache 2.0 permits commercial use, modification and redistribution, provided the license and copyright notices and any NOTICE file stay attached and significant changes are stated, and contributors grant patent rights. Two checks before committing: the page lists no token context length, so plan capacity in concurrent audio streams, and the Index shows no per-token host price for this model, so $1.85 an hour is the only benchmark.
Model Card
By OpenAI, published under apache-2.0, revision 973afd24965f.
Whisper
Whisper is a pre-trained model for automatic speech recognition (ASR) and speech translation. Trained on 680k hours of labelled data, Whisper models demonstrate a strong ability to generalise to many datasets and domains without the need for fine-tuning.
Whisper was proposed in the paper Robust Speech Recognition via Large-Scale Weak Supervision by Alec Radford et al from OpenAI. The original code repository can be found here.
Disclaimer: Content for this model card has partly been written by the Hugging Face team, and parts of it were copied and pasted from the original model card.
Model details
Whisper is a Transformer based encoder-decoder model, also referred to as a sequence-to-sequence model. It was trained on 680k hours of labelled speech data annotated using large-scale weak supervision.
Configuration
- Architecture
- WhisperForConditionalGeneration
- Layers
- 12
- Vocabulary size
- 51,865
- Stored precision
- float32
- Model type
- whisper
Identity and Version
- Repository
- openai/whisper-small
- Publisher
- OpenAI
- Task
- Speech recognition
- Modality
- Audio
- Library
- transformers
- Parameters
- 242M parameters
- Languages
- en, zh, de, es, ru, ko, fr, ja
- Revision
- 973afd24965f72e36ca33b3055d56a652f456b4d
- First published
- 2022-09-26
- Last updated
- 2024-02-29
Files and Weights
16 files, 3.9 GB in total. The weights are 4 files totalling 3.9 GB in bin, h5, msgpack, safetensors.
Weights4 files · 3.9 GB
Configuration6 files · 280.3 KB
Tokenizer4 files · 4.1 MB
Documentation1 file · 20.2 KB
Repository1 file · 1.4 KB
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| flax_model.msgpack | Weights | 967.0 MB | c375de483cc3 |
| model.safetensors | Weights | 967.0 MB | 1d7734884874 |
| pytorch_model.bin | Weights | 967.1 MB | ea40d8f6c99c |
| tf_model.h5 | Weights | 967.5 MB | c3b6ee117c26 |
| added_tokens.json | Configuration | 34.6 KB | — |
| config.json | Configuration | 2.0 KB | — |
| generation_config.json | Configuration | 3.9 KB | — |
| normalizer.json | Configuration | 52.7 KB | — |
| preprocessor_config.json | Configuration | 185.0 KB | — |
| special_tokens_map.json | Configuration | 2.2 KB | — |
| README.md | Documentation | 20.2 KB | — |
| .gitattributes | Repository | 1.4 KB | — |
| merges.txt | Tokenizer | 493.9 KB | — |
| tokenizer.json | Tokenizer | 2.5 MB | — |
| tokenizer_config.json | Tokenizer | 282.7 KB | — |
| vocab.json | Tokenizer | 835.5 KB | — |
License and Download
- License
- apache-2.0
- Access
- Open weights, no gate
- Download size
- 3.9 GB
Download from OpenAI
Released by OpenAI through its official repository on Hugging Face. Read the license.
Built From
- Described by arXiv:2212.04356
Evaluations
Each result is shown as reported, with the conditions its reporter stated. None is a SAVRN measurement. A comparison lines two results up only when their configuration, unit and setup are all stated and identical.
| Benchmark | Conditions | Result | Reported by | Revision | Date |
|---|---|---|---|---|---|
| Common Voice 11.0 | Configuration hiTask Automatic Speech RecognitionMetric Test WERComparison conditions not established | 87.3 | openai Publisher reported |
Evaluated revision not stated | — |
| Common Voice 13.0 | Configuration dvTask Automatic Speech RecognitionMetric WerComparison conditions not established | 125.698 | openai Publisher reported |
Evaluated revision not stated | — |
| LibriSpeech (clean) | Configuration cleanTask Automatic Speech RecognitionMetric Test WERComparison conditions not established | 3.43221 | openai Publisher reported |
Evaluated revision not stated | — |
| LibriSpeech (other) | Configuration otherTask Automatic Speech RecognitionMetric Test WERComparison conditions not established | 7.6283 | openai Publisher reported |
Evaluated revision not stated | — |
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 3.9 GB |
| 16-bit | 0.5 GB |
| 8-bit | 0.2 GB |
| 4-bit | 0.1 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Compare whisper-small
Questions About whisper-small
How much GPU memory does whisper-small need?
About 0.6 GB at 16-bit and 0.1 GB at 4-bit: the weights (242M parameters) plus a working margin. A long context needs more.
What is the cheapest GPU to run whisper-small on?
At 16-bit, 1x MI300X from $1.85 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.
Can I use whisper-small commercially?
Yes. whisper-small is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.
Similar Models
This Python package provides an efficient way to perform forced alignment between text and audio using Hugging Face's pretrained models. it also features an improved implementation to use much less memory than TorchAudio forced alignment API. The model checkpoint uploaded here is a conversion from torchaudio to HF Transformers for the MMS-300M checkpoint trained on forced alignment dataset
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the private datasets in 2 stages - firstly was fine-tuned on a small dataset with good samples Then the obtained model was fine-tuned on a large dataset with the small good dataset, with various samples from different sources, and with an unlabeled dataset that was weakly labeled using a previously trained model. (weakly labeled data wasn't used in validation set) on small dataset on large dataset on small dataset on large dataset The following hyperparameters were used during training: - learningrate: 0.0003 - trainbatchsize: 8 - evalbatchsize: 8 - distributedtype: multi-GPU - numdevices: 2 - gradientaccumulationsteps: 4…
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the FTSpeech dataset, being a dataset of 1,800 hours of transcribed speeches from the Danish parliament. The model achieves the following WER scores (lower is better): The use of this model needs to adhere to this license from the Danish Parliament.
You can test this model online with the Space for Romanian Speech Recognition The model ranked TOP-1 on Romanian Speech Recognition during HuggingFace's Robust Speech Challenge: This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the Common Voice 8.0 - Romanian subset dataset, with extra training data from Romanian Speech Synthesis dataset. Without the 5-gram Language Model optimization, it achieves the following results on the evaluation set (Common Voice 8.0, Romanian subset, test split): The architecture is based on facebook/wav2vec2-xls-r-300m with a speech recognition CTC head and an added 5-gram language model (using pyctcdecode and kenlm) trained on the Romanian…
ATTENTION! Metrics (float16) using evaluate library with batchsize=1