DeepSeek Strikes Again With AI Image Generator Janus-Pro
In the weekend’s new AI model virus success, Deepseek Has returned to another press conference-this time focused on AI image generation.
This Chinese startup has released a image generation model called Janus-PRO, which aims to conduct with American competitors From 3 and Stable diffusionEssence It is said that the performance of multi -model models that generate images in the text prompts are better than competitive services in the fields of image quality and accuracy.
After this release, after the release of Deepseek’s R1 model, the model caused waves with its lightning, highly logical response, and a small part of the cost of American model costs. According to reports, the model runs on the less advanced NVIDIA chip and proposes the question of how China competes without using frontier American technology. The application is recent More than Changpt Download on the Apple App Store.
Back release signal China has achieved basic promotion in the growing AI arms race competition. At the same time, last week, President Donald Trump announced a new AI infrastructure initiative to cooperate with OPENAI and other technology companies to guarantee up to $ 500 million.
The timing is also in line with the review of Chinese technology companies, and the tension situation has become higher and higher. Tiktok’s data privacy problemEssence
Read moreThe Deepseek is a hot AI application. Don’t be too attached
Janus-Pro is currently Download Hug the facehole on the AI ​​developer platform.
exist introduce Deepseek said on the download page: “Janus-PRO exceeds the previous unified model, matches or exceeds the performance of a model-specific model. Simple, high flexibility and the effectiveness of Janus-PRO make it a powerful candidate for the next generation “Unified Multi -Model”.
The size of this model ranges from 1 billion to 7 billion parameters, which is a key factor in solving the ability to solve the problem.
The company claims that Janus-PRO is the “new automatic regression framework”. It solves the previous challenges by separating the steps of analyzing and generating images. At the same time, a single unified system is still used to process all content.
DeepSeek wrote: “Decumbling not only reduces the conflict between the visual encoder in understanding and power generation, but also enhances the flexibility of the framework.”
