alt="drawing" width="600"/> If you already know T5, FLAN-T5 is just better at everything. For the same number of parameters, these models have been fine-tuned on more than 1000 additional tasks covering also more languages.
Runs On
What it takes to serve flan-t5-small (77M 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 | 0.2 GB | 0.2 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 Sep 18, 2026.
SAVRN's Notes on flan-t5-small
Do not size the card by the download. The repository weighs about 1.37 GB across 12 files because it ships in four formats, safetensors, pytorch, jax and tf, but the copy you load at 16-bit needs 0.2 GB, and 8-bit halves that to 0.1 GB. So this 77 million parameter T5ForConditionalGeneration model fits any accelerator you own; the cheapest setup we price, one MI300X with 192 GB at $1.85 an hour, leaves nearly all of that memory free.
Apache 2.0 permits commercial deployment, modification and redistribution, provided the notices stay and significant changes are stated, with a patent grant included. Check two things: no context length is stated, so measure the input size you need yourself, and the last update was October 10, 2023. The training relations name gsm8k, aqua_rat, esnli, lambada and deepmind/code_contests, and Google describes tuning on more than 1,000 additional tasks beyond T5.
Model Card
By Google, published under apache-2.0, revision 0fc9ddf78a1e.
Model Card for FLAN-T5 small
Table of Contents
- TL;DR
- Model Details
- Usage
- Uses
- Bias, Risks, and Limitations
- Training Details
- Evaluation
- Environmental Impact
- Citation
- Model Card Authors
TL;DR
If you already know T5, FLAN-T5 is just better at everything. For the same number of parameters, these models have been fine-tuned on more than 1000 additional tasks covering also more languages. As mentioned in the first few lines of the abstract :
Flan-PaLM 540B achieves state-of-the-art performance on several benchmarks, such as 75.2% on five-shot MMLU. We also publicly release Flan-T5 checkpoints,1 which achieve strong few-shot performance even compared to much larger models, such as PaLM 62B. Overall, instruction finetuning is a general method for improving the performance and usability of pretrained language models.
Disclaimer: Content from this model card has been written by the Hugging Face team, and parts of it were copy pasted from the T5 model card.
Model Details
Model Description
Configuration
- Architecture
- T5ForConditionalGeneration
- Vocabulary size
- 32,128
- Model type
- t5
Identity and Version
- Repository
- google/flan-t5-small
- Publisher
- Task
- Not stated by the source
- Modality
- Other
- Library
- transformers
- Parameters
- 77M parameters
- Languages
- en, fr, ro, de
- Revision
- 0fc9ddf78a1e988dac52e2dac162b0ede4fd74ab
- First published
- 2022-10-21
- Last updated
- 2023-10-10
Files and Weights
12 files, 1.4 GB in total. The weights are 4 files totalling 1.4 GB in bin, h5, msgpack, safetensors.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| flax_model.msgpack | Weights | 307.9 MB | 46dc0494fa27 |
| model.safetensors | Weights | 307.9 MB | 495fa51e2046 |
| pytorch_model.bin | Weights | 307.9 MB | 4a8c0174fa3e |
| tf_model.h5 | Weights | 439.8 MB | 9c951cf31dec |
| config.json | Configuration | 1.4 KB | — |
| generation_config.json | Configuration | 147 B | — |
| special_tokens_map.json | Configuration | 2.2 KB | — |
| README.md | Documentation | 10.8 KB | — |
| .gitattributes | Repository | 1.4 KB | — |
| spiece.model | Tokenizer | 791.7 KB | d60acb128cf7 |
| tokenizer.json | Tokenizer | 2.4 MB | — |
| tokenizer_config.json | Tokenizer | 2.5 KB | — |
License and Download
- License
- apache-2.0
- Access
- Open weights, no gate
- Download size
- 1.4 GB
Released by Google through Kaggle. Read the license.
Built From
- Described by arXiv:1910.09700
- Described by arXiv:2210.11416
- Trained on (disclosed) aqua_rat
- Trained on (disclosed) deepmind/code_contests
- Trained on (disclosed) djaym7/wiki_dialog
- Trained on (disclosed) esnli
- Trained on (disclosed) gsm8k
- Trained on (disclosed) lambada
- Trained on (disclosed) qed
- Trained on (disclosed) quasc
- Trained on (disclosed) svakulenk0/qrecc
- Trained on (disclosed) taskmaster2
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 1.4 GB |
| 16-bit | 0.2 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.
Questions About flan-t5-small
How much GPU memory does flan-t5-small need?
About 0.2 GB at 16-bit and 0 GB at 4-bit: the weights (77M parameters) plus a working margin. A long context needs more.
What is the cheapest GPU to run flan-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 flan-t5-small commercially?
Yes. flan-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.