source languages: fr; target languages: en; OPUS readme: fr-en; dataset: opus; model: transformer-align; pre-processing: normalization + SentencePiece.
t5-small is an open-weight model for translation from T5 community, released under Apache License 2.0. It has 61M parameters. At 16-bit it needs about 0.1 GB of GPU memory, which fits on 1x MI300X from $1.85 an hour, at the lowest prices in the SAVRN Index. It draws 24.4M downloads a month.
The developers of the Text-To-Text Transfer Transformer (T5) write: T5-Small is the checkpoint with 60 million parameters. The developers write in a blog post that the model: See the blog post and research paper for further details.
Parameters61M
Context—
Weights2.2 GB
Licenseapache-2.0
AccessOpen weights
Monthly Downloads24.4M
Runs On
What it takes to serve t5-small (61M 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 | 0.1 GB | 0.1 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 8-bit | 0.1 GB | 0.1 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 4-bit | 0.0 GB | 0.0 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 Oct 9, 2026.
t5-small on every accelerator the SAVRN Index prices, at every precision
SAVRN's Notes on t5-small
The 4-bit row reads 0.0 GB and the 16-bit row 0.1 GB, so quantizing buys nothing worth the trouble. The cheapest setup we list is a 192 GB MI300X at $1.85 an hour, which makes the only hardware question what else that card does. What you get is the 61M-parameter checkpoint of the text-to-text transfer transformer, pretrained on C4 with a mix of unsupervised and supervised tasks, labeled for translation.
Apache 2.0 means you can fine-tune it, ship the result and sell it, keeping the notices and stating what you changed. No context length is listed, so measure your longest input against the model yourself. The seven papers listed for it are task datasets, acceptability judgments, semantic textual similarity, inference, SQuAD, word in context, reading comprehension and yes-or-no questions, the trail for what the supervised side covered. The last update was 2023-06-30.
Model Card
By T5 community, published under apache-2.0, revision df1b051c4962.
Model Card for T5 Small
Table of Contents
- Model Details
- Uses
- Bias, Risks, and Limitations
- Training Details
- Evaluation
- Environmental Impact
- Citation
- Model Card Authors
- How To Get Started With the Model
Model Details
Model Description
The developers of the Text-To-Text Transfer Transformer (T5) write:
With T5, we propose reframing all NLP tasks into a unified text-to-text-format where the input and output are always text strings, in contrast to BERT-style models that can only output either a class label or a span of the input. Our text-to-text framework allows us to use the same model, loss function, and hyperparameters on any NLP task.
T5-Small is the checkpoint with 60 million parameters.
Configuration
- Architecture
- T5ForConditionalGeneration
- Vocabulary size
- 32,128
- Model type
- t5
Identity and Version
- Repository
- google-t5/t5-small
- Publisher
- T5 community
- Task
- Translation
- Modality
- Text
- Library
- transformers
- Parameters
- 61M parameters
- Languages
- en, fr, ro, de
- Revision
- df1b051c49625cf57a3d0d8d3863ed4d13564fe4
- First published
- 2022-03-02
- Last updated
- 2023-06-30
Files and Weights
20 files, 2.2 GB in total. The weights are 13 files totalling 2.2 GB in bin, h5, msgpack, onnx, ot, safetensors.
Weights13 files · 2.2 GB
Configuration2 files · 1.4 KB
Tokenizer3 files · 2.2 MB
Documentation1 file · 8.5 KB
Repository1 file · 537 B
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| flax_model.msgpack | Weights | 242.0 MB | b143e13ccb73 |
| model.safetensors | Weights | 242.0 MB | bd944e5f1b3a |
| onnx/decoder_model.onnx | Weights | 232.5 MB | 2256c6365e82 |
| onnx/decoder_model_merged.onnx | Weights | 232.6 MB | f6d2874dafd2 |
| onnx/decoder_model_merged_quantized.onnx | Weights | 58.7 MB | af6a06ea8488 |
| onnx/decoder_model_quantized.onnx | Weights | 58.4 MB | a1b7a03fd50c |
| onnx/decoder_with_past_model.onnx | Weights | 219.9 MB | 2eb7790f56c4 |
| onnx/decoder_with_past_model_quantized.onnx | Weights | 55.2 MB | 30e4b22fc986 |
| onnx/encoder_model.onnx | Weights | 141.4 MB | 3ebb633a3a81 |
| onnx/encoder_model_quantized.onnx | Weights | 35.5 MB | 5124770e67bf |
| pytorch_model.bin | Weights | 242.1 MB | dd8c1c79a54c |
| rust_model.ot | Weights | 242.1 MB | 636b17628823 |
| tf_model.h5 | Weights | 242.3 MB | fc9d2ec530ef |
| config.json | Configuration | 1.2 KB | — |
| generation_config.json | Configuration | 147 B | — |
| README.md | Documentation | 8.5 KB | — |
| .gitattributes | Repository | 537 B | — |
| spiece.model | Tokenizer | 791.7 KB | — |
| tokenizer.json | Tokenizer | 1.4 MB | — |
| tokenizer_config.json | Tokenizer | 2.3 KB | — |
License and Download
- License
- apache-2.0
- Access
- Open weights, no gate
- Download size
- 2.2 GB
Download from T5 community
Released by T5 community through its official repository on Hugging Face. Read the license.
Built From
- Described by arXiv:1606.05250
- Described by arXiv:1704.05426
- Described by arXiv:1708.00055
- Described by arXiv:1805.12471
- Described by arXiv:1808.09121
- Described by arXiv:1810.12885
- Described by arXiv:1905.10044
- Described by arXiv:1910.09700
- Trained on (disclosed) c4
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 2.2 GB |
| 16-bit | 0.1 GB |
| 8-bit | 0.1 GB |
| 4-bit | 0.0 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Built on This Model
- Derived frombasic-chat-model
Compare t5-small
Questions About t5-small
How much GPU memory does t5-small need?
About 0.1 GB at 16-bit and 0 GB at 4-bit: the weights (61M parameters) plus a working margin. A long context needs more.
What is the cheapest GPU to run t5-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 t5-small commercially?
Yes. t5-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
基于标准 Transformer（Post-LN + 固定 sin-cos 位置编码）的英译中模型， 由本工程用 HuggingFace Seq2SeqTrainer 训练得到（checkpoint：trained with Seq2SeqTrainer (standard HF)）。 本仓库是标准 HF 仓库：包含 config.json + automap，代码随仓库分发， 因此任何人在 HF 上 trustremotecode=True 即可直接加载使用，无需本工程的其它文件。 逐步自回归推理，未实现 KV Cache（usecache=false），长句偏慢。 生成质量以 beam search（numbeams=3）优于贪心。 请按自身场景补充 LICENSE。
source languages: nl; target languages: en; OPUS readme: nl-en; dataset: opus; model: transformer-align; pre-processing: normalization + SentencePiece.
This is the model card of NLLB-200's distilled 600M variant. Here are the metrics for that particular checkpoint. - Information about training algorithms, parameters, fairness constraints or other applied approaches, and features. The exact training algorithm, data and the strategies to handle data imbalances for high and low resource languages that were used to train NLLB-200 is described in the paper. - Paper or other resource for more information NLLB Team et al, No Language Left Behind: Scaling Human-Centered Machine Translation, Arxiv, 2022 - Where to send questions or comments about the model: https://github.com/facebookresearch/fairseq/issues • Model performance measures: NLLB-200…
source languages: en; target languages: ru; OPUS readme: en-ru; dataset: opus; model: transformer-align; pre-processing: normalization + SentencePiece.
This model can be used for translation and text-to-text generation. CONTENT WARNING: Readers should be aware this section contains content that is disturbing, offensive, and can propagate historical and current stereotypes. Significant research has explored bias and fairness issues with language models (see, e.g., Sheng et al. (2021) and Bender et al. (2021)). Further details about the dataset for this model can be found in the OPUS readme: en-de