PixArt-Σ
by PixArt-alpha (HKUST / Huawei Noah's Ark Lab) · open-source · Last verified 2026-03-17
PixArt-Σ is a high-efficiency text-to-image diffusion transformer model that achieves competitive quality with models 10× larger by leveraging a novel weak-to-strong training strategy and high-quality data curation. It supports 4K resolution output and is fully open-source, making it an attractive choice for researchers and developers seeking quality without compute overhead.
https://pixart-alpha.github.io/PixArt-sigma-project/ ↗C
C—Below Average
Adoption: CQuality: AFreshness: B+Citations: C+Engagement: F
Specifications
- License
- Apache 2.0
- Pricing
- open-source
- Capabilities
- text-to-image, 4k-resolution-output, high-quality-generation, fine-tuning, efficient-inference
- Integrations
- huggingface, diffusers, comfyui
- Use Cases
- high-resolution-art, research, commercial-imagery, fine-tuning-experiments, print-quality-output
- API Available
- Yes
- Parameters
- 600M
- Context Window
- N/A
- Modalities
- text, image
- Training Cutoff
- 2024
- Tags
- image-generation, text-to-image, open-source, diffusion-transformer, efficient
- Added
- 2026-03-17
- Completeness
- 100%
Index Score
48.5Adoption
44
Quality
82
Freshness
72
Citations
58
Engagement
0