One-two AI Punch: DeepSeek's image generator follows its edition of the country model


How to Launching a new AI model that shook the entire industry wasn't enough, Chinese startup DeepSeek followed this week with release Ai's image generator claims it claims It provides “significant achievements in both multimodal understanding and opportunities for monitoring image instructions to image”.

The new model for image generation is called Janus-Proand aims to compete with us rivals like Of 3 And Stable diffusion. The new model claims to surpass its competition in areas such as image quality and accuracy.

Ai Atlas Art Badge Label

The launch of Janus-Pro came just days after the announcement of Model R1 on DeepSeekWhat has done waves with its lightning fast, highly logical reaction and as soon as possible with some of the costs of American models.

The DeepSeek model is reportedly working on less advanced NVIDIA chips, asking questions about how China competes without access to top US technology. The iOS app has External chatgpt In the downloads of the Apple App Store recently, it is still a free application No. 1 from January 31st.

Announcements on the back of the back signal the pressure of China to acquire the footage of the growing AI arming race. Meanwhile, last week, President Donald Trump has released a new AI infrastructure initiative, promising to $ 500 million in partnership with Openii and other tech companies.

Watch this: What is DeepSeek AI? Everything to know about the popular new AI

The release of R1 and Janus-Pro also coincides with the increased review of Chinese tech companies, with tension already high Concern about Tiktok data privacy.

In an introduction to its download page, DeepSeek says: “Janus-Pro exceeds the previous unified model and fits or exceeds the performance of task-specific models. The simplicity, high flexibility and effectiveness of Janus-Pro made a strong candidate for the next generation multimodal models. “

The model ranges from 1 billion to 7 billion parameters, a key factor in its troubleshooting capabilities.

The company calls it Janus-Pro as a “Roman autorgredient framework” that solves previous challenges by taking steps to analyze and generate images, while still using a unique, unified system for processing everything.

“The separation not only mitigates the conflict between the roles of the visual encoder in understanding and the generation, but it improves the flexibility of the frame,” wrote dipssee.

The Janus-Pro user's response has been mixed so far, with some red -ions that claim Pictures resemble the efforts of his competitors from past years. To get a sense of how Janus-Pro compares to other Ai's image generators, check this review of the performance between Chatgpt 4o, Qwen 2.5 and Janus-Pro of YouTuber Ejack Yao.

Janus-Pro right now Available for download The AI ​​developer platform hugs the face.





Source link

Leave a Reply

Your email address will not be published. Required fields are marked *