SAVRN Model Hub · Comparisons
convnextv2_nano.fcmae_ft_in1k vs efficientnet_b3.ra2_in1k
Convnextv2_nano.fcmae_ft_in1k has 16M parameters and efficientnet_b3.ra2_in1k has 12M parameters; convnextv2_nano.fcmae_ft_in1k is released under Creative Commons Attribution-NonCommercial 4.0 and efficientnet_b3.ra2_in1k under Apache License 2.0; at 16-bit, convnextv2_nano.fcmae_ft_in1k needs about 0 GB (1x MI300X from $1.85 an hour) and efficientnet_b3.ra2_in1k about 0 GB (1x MI300X from $1.85 an hour).
| Field | convnextv2_nano.fcmae_ft_in1k timm/convnextv2_nano.fcmae_ft_in1k | efficientnet_b3.ra2_in1k timm/efficientnet_b3.ra2_in1k |
|---|---|---|
| Publisher | PyTorch Image Models | PyTorch Image Models |
| Task | Image classification | Image classification |
| Modality | Image | Image |
| Parameters, as reported | 16M parameters | 12M parameters |
| Architecture | Not stated | Not stated |
| Library | timm | timm |
| Context length | Not stated | Not stated |
| Repository size | 125.1 MB | 98.8 MB |
| Artifact formats | safetensors, pytorch | safetensors, pytorch |
| License | cc-by-nc-4.0 | apache-2.0 |
| Access | Open weights, no gate | Open weights, no gate |
| Memory at 16-bit (weights and margin) | 0 GB | 0 GB |
| Cheapest GPUs at 16-bit, per hour | 1x MI300X, $1.85 | 1x MI300X, $1.85 |
| Memory at 4-bit (weights and margin) | 0 GB | 0 GB |
| Cheapest GPUs at 4-bit, per hour | 1x MI300X, $1.85 | 1x MI300X, $1.85 |
| Revision viewed | fd83d9c98615 | 0366a7551862 |
| Downloads reported by the hub | 165.3k | 1M |
| Last observed | 2026-09-18 | 2026-10-02 |
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 efficientnet_b3.ra2_in1k
Twelve million parameters trained with the RandAugment RA2 recipe on ImageNet-1k, and the whole thing rounds to 0.0 GB at every precision on our table against a 192 GB MI300X at $1.85 an hour. Nobody rents a datacenter GPU for this; the honest deployment target is CPU or the smallest edge accelerator you own, and quantization is pointless at this size. This is a timm 'B' recipe, evolved from the ResNet Strikes Back line, so read it as a reference implementation rather than a frontier result.
Apache 2.0 permits commercial use and redistribution with the standard notice and patent terms. Treat it as a baseline: the same timm family lists ConvNeXtV2, DenseNet-121 and RepVGG neighbors that beat it on accuracy-per-dollar for many tasks, so benchmark before defaulting to EfficientNet out of habit. It draws 1.1M downloads a month and stays popular mostly as a cheap pretraining starting point and a stable comparison target.
Questions
Which is larger, convnextv2_nano.fcmae_ft_in1k or efficientnet_b3.ra2_in1k?
convnextv2_nano.fcmae_ft_in1k (16M parameters) is larger than efficientnet_b3.ra2_in1k (12M parameters), by the parameter counts their publishers report.
Which is cheaper to run, convnextv2_nano.fcmae_ft_in1k or efficientnet_b3.ra2_in1k?
At 4-bit, convnextv2_nano.fcmae_ft_in1k fits on 1x MI300X from $1.85 an hour and efficientnet_b3.ra2_in1k on 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.
Can I use convnextv2_nano.fcmae_ft_in1k commercially?
Not without separate permission. convnextv2_nano.fcmae_ft_in1k is released under Creative Commons Attribution-NonCommercial 4.0. CC BY-NC 4.0 permits sharing and adapting with credit for non-commercial purposes only. Commercial use needs separate permission from the rights holder.
Can I use efficientnet_b3.ra2_in1k commercially?
Yes. efficientnet_b3.ra2_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.