Is PixAI Tagger 1.0 state of the art?Pending benchmarks.Announced Oct 22, 2026; no independent results yet.
PixAI's open anime image tagger, the model PixAI used to curate training data for its Tsubaki.3 image model. A fine-tuned SAM 3 ViTDet-style image encoder (1008 × 1008 input, 32 blocks) feeds a classification head that predicts 30,877 tags for characters, clothing, styles and more. 486.3M parameters; weights on Hugging Face under Apache-2.0. Not an LLM.
The details so far
- Lab
- PixAI
- Announced
- Oct 22, 2026
- API
- Not yet
- Open weights
- Published
- Parameters
- 486.3M
- Task
- Anime image tagging (multi-label classification)
- Backbone
- Fine-tuned SAM 3 ViTDet-style image encoder, 1008 × 1008 input, 32 blocks
- Vocabulary
- 30,877 tags across six categories, up from 13,461 in v0.9; knowledge cutoff May 2026
- Use at PixAI
- Used throughout Tsubaki.3's data curation (lab post)
- Weights
- Hugging Face (pixai-labs/pixai-tagger-v1.0)
- License
- Apache-2.0 (model card)
Tags marked unsure are not stated on the lab's own pages.
What would let us call it
Verdicts use only independent boards we already track. Two consecutive snapshots are required before any crown.
- Publicly announced with a source
- Usable by the public (API or weights)
- Listed on an independent board we track
- Enough results for a rank in a category
- Holds across two snapshots (crowns only)
What the lab claims
Self-reported. Shown for context; not used in any verdict.
Lab claim: highest General-tag micro F1 (0.6660), macro F1 and mAP among the eight anime taggers in PixAI's shared-tag benchmark.
Sources
REPOPixAI Tagger v1.0 model card on Hugging Facehuggingface.co/pixai-labs/pixai-tagger-v1.0 · Oct 22, 2026LABPixAI announcement on Xx.com/PixAI_EN/status/2113284257061736470 · Oct 22, 2026Tracking continues until an independent board lists it.