Linum-AI has released P-JiT 21K, a text-to-image model developed using the P-JiT architecture. This model was trained on 610 million samples at 512px resolution and 50 million samples at 1024px resolution. The training process cost approximately $21,000 and incorporated a variant of the FDFO technique for post-training. AI
IMPACT Offers a new option for text-to-image generation, potentially impacting creative workflows and research in generative AI.
RANK_REASON Release of a new text-to-image model with details on training cost and architecture. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →