Model card
SoViT-400m/14.
Google DeepMindopen-source400M paramsVision Transformer (Shape-Optimized)
Shape-Optimized ViT derived from compute-optimal scaling laws. 400M params, patch14. Pre-trained on WebLI (contrastive + reconstruction head), finetuned on ImageNet-1K at 224px. NeurIPS 2023. Paper: arxiv:2305.13035.
§ 01 · Benchmarks
Every benchmark SoViT-400m/14 has a recorded score for.
| # | Benchmark | Area · Task | Metric | Value | Rank | Date | Source |
|---|---|---|---|---|---|---|---|
| 01 | ImageNet-1K | Computer Vision · Image Classification | top-1-accuracy | 90.3% | #3 | — | source ↗ |
Rank column shows this model’s position vs all other models scored on the same benchmark + metric (competitors after the slash). #1 in red means current SOTA. Sorted by rank, then newest result.
§ 02 · Strengths by area
Where SoViT-400m/14 actually performs.
§ 04 · Related models
Other Google DeepMind models scored on Codesota.
§ 05 · Sources & freshness
Where these numbers come from.
neurips-2023
1
result
0 of 1 rows marked verified.