SAVRN Model Hub · Comparisons
vit-base-patch16-224 vs vit_base_patch8_224.augreg2_in21k_ft_in1k
Vit-base-patch16-224 has 87M parameters and vit_base_patch8_224.augreg2_in21k_ft_in1k has 87M parameters; both are released under Apache License 2.0; at 16-bit, vit-base-patch16-224 needs about 0.2 GB (1x MI300X from $1.85 an hour) and vit_base_patch8_224.augreg2_in21k_ft_in1k about 0.2 GB (1x MI300X from $1.85 an hour).
| Field | vit-base-patch16-224 google/vit-base-patch16-224 | vit_base_patch8_224.augreg2_in21k_ft_in1k timm/vit_base_patch8_224.augreg2_in21k_ft_in1k |
|---|---|---|
| Publisher | PyTorch Image Models | |
| Task | Image classification | Image classification |
| Modality | Image | Image |
| Parameters, as reported | 87M parameters | 87M parameters |
| Architecture | ViTForImageClassification | Not stated |
| Library | transformers | timm |
| Context length | Not stated | Not stated |
| Repository size | 1.4 GB | 692.7 MB |
| Artifact formats | safetensors, pytorch, jax, tf | safetensors, pytorch |
| License | apache-2.0 | apache-2.0 |
| Access | Open weights, no gate | Open weights, no gate |
| Memory at 16-bit (weights and margin) | 0.2 GB | 0.2 GB |
| Cheapest GPUs at 16-bit, per hour | 1x MI300X, $1.85 | 1x MI300X, $1.85 |
| Memory at 4-bit (weights and margin) | 0.1 GB | 0.1 GB |
| Cheapest GPUs at 4-bit, per hour | 1x MI300X, $1.85 | 1x MI300X, $1.85 |
| Revision viewed | 3f49326eb077 | 907a22023d1c |
| Downloads reported by the hub | 7.1M | 275.1k |
| Last observed | 2026-09-18 | 2026-09-18 |
An evaluation row appears only where at least two of these models report the same benchmark with the same stated configuration, metric, unit and setup. Different evaluators stay named in each cell. Values are shown as reported: no unit conversion, no ranking.
SAVRN's Notes on vit-base-patch16-224
You will barely notice this one on a card. At 16-bit it needs 0.2 GB to run, and the cheapest setup we list, one MI300X with 192 GB at $1.85 an hour on demand, has room for it hundreds of times over. It sorts a 224 by 224 image into one of 1,000 ImageNet 2012 classes; Google pre-trained its 87M parameters on 14 million ImageNet-21k images. The accelerator question is images per hour and what else shares the card, never fit.
Apache 2.0 lets you fine-tune it to your own labels, run it commercially and redistribute it, provided the license and NOTICE file travel along and you state what you changed, with a patent grant included. Check that the 1,000 ImageNet labels cover your need; anything else means training a new head on your data. The weights date from September 2023 and are a conversion from the timm repository.
Questions
Which is larger, vit-base-patch16-224 or vit_base_patch8_224.augreg2_in21k_ft_in1k?
vit_base_patch8_224.augreg2_in21k_ft_in1k (87M parameters) is larger than vit-base-patch16-224 (87M parameters), by the parameter counts their publishers report.
Which is cheaper to run, vit-base-patch16-224 or vit_base_patch8_224.augreg2_in21k_ft_in1k?
At 4-bit, vit-base-patch16-224 fits on 1x MI300X from $1.85 an hour and vit_base_patch8_224.augreg2_in21k_ft_in1k on 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.
Can I use vit-base-patch16-224 commercially?
Yes. vit-base-patch16-224 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.
Can I use vit_base_patch8_224.augreg2_in21k_ft_in1k commercially?
Yes. vit_base_patch8_224.augreg2_in21k_ft_in1k 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.