DeepSeek, a fast-growing AI company, has introduced its latest suite of multimodal AI models under the Janus-Pro family, which it claims can outperform OpenAI’s DALL-E 3 in key benchmarks. These models, ranging from 1 billion to 7 billion parameters, are available for download on the AI platform Hugging Face and come with an MIT license, allowing for unrestricted commercial use.
The Janus-Pro models use a “novel autoregressive framework” designed for both image analysis and creation. According to DeepSeek, the largest model, Janus-Pro-7B, has shown superior performance on benchmarks such as GenEval and DPG-Bench, surpassing other popular models like PixArt-alpha, Emu3-Gen, and Stable Diffusion XL. Despite some limitations, such as image size restrictions (up to 384 x 384 resolution), the compact design of Janus-Pro continues to impress the AI community.
Also Read: DeepSeek Overtakes ChatGPT— Here’s Why It’s Gaining Massive Popularity
DeepSeek’s innovative approach has already captured attention, with its chatbot app topping the Apple App Store charts. The company is largely funded by High-Flyer Capital Management and is challenging the status quo in AI development, raising questions about global competition in AI and the increasing demand for AI chips.
