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-base (248M 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 flan-t5-base
October 2022 is the release date, July 2023 the last update, and it still pulls 1,381,699 downloads a month. Size is the reason. At 248 million parameters it needs 0.6 GB of memory at 16-bit, 0.3 GB at 8-bit and 0.1 GB at 4-bit, a rounding error on the 192 GB MI300X that is our cheapest priced card at $1.85 an hour, so it rides beside other work rather than owning a GPU. The architecture is T5ForConditionalGeneration, an encoder-decoder from Google tuned on more than 1,000 additional tasks; the training list in our file includes gsm8k, aqua_rat, esnli, lambada and code_contests.
Apache 2.0 covers commercial use, modification and redistribution with the notices kept and significant changes stated. Check two things before committing: no task and no context length are listed, so you establish both by testing on your own inputs, and the 4.16 GB download spans 12 files because it ships in safetensors, pytorch, jax and tf, so pull only the format your stack loads.
Model Card
By Google, published under apache-2.0, revision 7bcac572ce56.
Model Card for FLAN-T5 base
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-base
- Publisher
- Task
- Not stated by the source
- Modality
- Other
- Library
- transformers
- Parameters
- 248M parameters
- Languages
- en, fr, ro, de
- Revision
- 7bcac572ce56db69c1ea7c8af255c5d7c9672fc2
- First published
- 2022-10-21
- Last updated
- 2023-07-17
Files and Weights
12 files, 4.2 GB in total. The weights are 4 files totalling 4.2 GB in bin, h5, msgpack, safetensors.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| flax_model.msgpack | Weights | 990.3 MB | e08fabc4ea8c |
| model.safetensors | Weights | 990.3 MB | 1dfb70afdced |
| pytorch_model.bin | Weights | 990.4 MB | 521ad5575487 |
| tf_model.h5 | Weights | 1.2 GB | 8cf13641aeca |
| config.json | Configuration | 1.4 KB | — |
| generation_config.json | Configuration | 147 B | — |
| special_tokens_map.json | Configuration | 2.2 KB | — |
| README.md | Documentation | 13.4 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
- 4.2 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 | 4.2 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.
Built on This Model
- Derived fromflan-t5-base-squad2
Questions About flan-t5-base
How much GPU memory does flan-t5-base need?
About 0.6 GB at 16-bit and 0.1 GB at 4-bit: the weights (248M parameters) plus a working margin. A long context needs more.
What is the cheapest GPU to run flan-t5-base 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-base commercially?
Yes. flan-t5-base 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.