DeepSeek unveils AI model claimed to outperform OpenAI’s DALL-E 3 and Stable Diffusion

Chinese startup DeepSeek has launched Janus-Pro, a new AI model that generates detailed images from text input. The model builds on the original Janus platform and has gained attention for outperforming established competitors like OpenAI’s DALL-E 3 and Stability AI’s Stable Diffusion, according to benchmark rankings in the company’s latest technical report.

DeepSeek claims Janus-Pro surpasses its rivals through enhanced training techniques, improved data quality, and an expanded model size. Incorporating 72 million high-quality synthetic images and real-world data, the model produces images with better stability and finer detail compared to existing models.

With an architecture featuring up to 7 billion parameters, Janus-Pro claims that it significantly improves training speed, accuracy, and comprehension of text-to-image tasks. These advancements allow the model to deliver visually compelling results even for complex prompts, strengthening DeepSeek’s position in the competitive AI image generation space.

Janus-Pro’s launch follows the success of DeepSeek’s chatbot, R1, powered by the DeepSeek-V3 model. The chatbot has become the top-rated free app on Apple’s US App Store, highlighting the startup’s growing influence in the AI industry. DeepSeek’s innovation is reshaping perceptions of Chinese AI firms, with its models gaining global recognition.

The introduction of Janus-Pro shows the growing competition in AI development, particularly in the open-source sector. Its strong performance has drawn comparisons to DALL-E 3, backed by Microsoft and OpenAI, as well as Stability AI’s Stable Diffusion. As DeepSeek sets new benchmarks, it puts pressure on major players like OpenAI to keep pace.

Meanwhile, the success of companies like DeepSeek is reshaping the AI landscape. Tech stocks such as Nvidia and Oracle dipped following the announcement, possibly indicating growing market competition and investor interest in emerging AI leaders.