Fresh to viral success of the new AI model over the weekend, Deepseek has returned to another launch – this time dedicated to the AI image generation.
Chinese startup released an image generation model called Janus-Pro, which aims to take US rivals From 3 and Stable diffusion. The multimodal model, which generates images from text signals, is said to be outperform competing services in areas such as image quality and accuracy.
This launch complies with the release of the R1 R1 model of DeepSeek, which makes waves along with fast lightning, highly logical responses, and for training faster and at a part of the cost of models of the models of the models of the models US. The model has been reported running on the less advanced Nvidia chips, which raises questions about how China competes without access to US cutting technology. The app recently Outpaced chatgpt Downloads to the Apple App Store.
The back-to-back releases China's pushing signal to get a walk to the growing AI weapon race. Meanwhile, last week, President Donald Trump announced a new AI infrastructure initiative, which promised up to $ 500 million in collaboration with Openai and other tech companies.
The timing also coincides with the increasing investigation of Chinese tech companies, with tensions high Tiktok data privacy concerns.
Read more: Deepseek is the hot AI app. Don't be too attached to it
Janus-Pro is currently Available to download In the AI developer platform that hugs the face.
In a Introduction On the download page, Deepseek says: “Janus-pro exceeds the previous unified model and match or exceeds the performance of models specific work. Simplicity, high flexibility, and effectiveness of Janus -pro is making a strong candidate for the next generation a multimodal models.
The model ranges from size from 1 billion to 7 billion parameters, a major factor in problem solving capabilities.
The company calls Janus-Pro which is a “novel Autoregressive Framework” that solves past challenges by separating steps for reviewing and developing images, while still using a single, unified system to process Everything.
“Decay not only eases the conflict between the visual encoder duties in understanding and generation but also enhances the flexibility of the plot,” Deepseek wrote.
