Chinese AI company DeepSeek has released a new family of AI models called Janus-Pro, with capabilities in both image analysis and creation. The models, ranging from 1 billion to 7 billion parameters, are available for download on the Hugging Face platform under an MIT license, allowing unrestricted commercial use.
According to DeepSeek, the largest model in the family, Janus-Pro-7B, outperforms several established competitors including OpenAI’s DALL-E 3 and Stability AI’s Stable Diffusion XL on two AI evaluation benchmarks: GenEval and DPG-Bench. The company describes Janus-Pro as a “novel autoregressive framework” designed to balance performance with computational efficiency, though most models in the family are currently limited to analyzing images with resolutions up to 384 x 384 pixels.
The release comes at a significant moment in the AI industry, as DeepSeek has recently gained prominence following the success of its chatbot application and language models. The company, primarily funded by High-Flyer Capital Management, has drawn attention from Wall Street analysts and technology experts for its compute-efficient training techniques. The launch of Janus-Pro has coincided with market reactions affecting U.S. AI companies, raising questions about global AI competition and leadership.
Sources: TechCrunch, VentureBeat