Tech Desk 
Jan 28, 2025
DeepSeek has officialy launched of its comprehensive suite of multimodal AI models, known as Janus-Pro. It is particularly designed to analyse and generate images, aiming to outdo OpenAI’s DALL-E 3 in several AI evaluation benchmarks.
What is Janus-Pro?
The recently launched Janus-Pro models are now available to install on the AI development platform Hugging Face. These models belong to the latest family, with models ranging from one-to-seven billion parameters.
Despite their slim sizes, DeepSeek particularly designed the Janus-Pro models to achieve excellence at both image analysis and creation. It licenses the models under the MIT license, enabling unrestricted commercial use. The image displays the capabilities of Jnaus and Janus Pro.
Janus-Pro performance
The largest model in the Janus-Pro family, Janus-Pro-7B, has allegedly exceeded renowned AI models such as OpenAI’s DALL-E 3, PixArt-alpha, Emu3-Gen, and Stability AI’s Stable Diffusion XL on two AI benchmarks: GenEval and DPG-Bench.
What sets Janus-Pro apart from previous models is its improves capabilities. It combines both versatility and flexibility across different apps, marking a groundbreaking development to reshape the AI industry.
The launch of the Janus-Pro models comes on the heels of the rapid rise of DeepSeek’s chatbot app, which sharply rose to the top of the Apple App Store charts.
Their continuous progress in both product development and market influence suggests that DeepSeek is set to revolutionise the AI industry in the coming years.
