DeepSeek Launches Janus Pro 7B: An Open-Source AI Model That Outperforms DALL-E 3

DeepSeek's Open-Source AI Evolution

Chinese AI firm DeepSeek has made waves once again by launching Janus Pro 7B, an advanced open-source image generation model. This release comes just days after the debut of DeepSeek-R1, a reasoning-focused AI model. The company claims Janus Pro 7B surpasses OpenAI's DALL-E 3 in multiple benchmark tests, reinforcing its commitment to fully open-source AI development. Like its predecessors, the model is available under a permissive license for both academic and commercial applications.

DeepSeek Janus Pro 7B: Features & Advancements

According to a Hugging Face listing, Janus Pro 7B is the successor to the Janus and Janus Pro 1B models, featuring major upgrades to its architecture. DeepSeek describes it as an autoregressive framework designed for multimodal understanding and generation. The model leverages a SigLIP-L vision encoder for image processing and a downsample rate of 16 tokeniser for efficient output generation.

To enhance performance, DeepSeek has optimized the model by decoupling visual encoding into separate pathways, while still utilizing a unified transformer architecture. These enhancements make Janus Pro 7B a powerful tool for AI-generated imagery.

Janus Pro 7B Benchmark Performance: Better Than DALL-E 3?

DeepSeek reports that Janus Pro 7B achieved:✅ 80% on GenEval✅ 84.2 on DPG-Bench

These scores outperform OpenAI’s DALL-E 3 and Stable Diffusion, solidifying Janus Pro 7B as a leading open-source AI model. However, independent third-party testing will provide a more comprehensive evaluation in the coming days.

The model is now available for download on GitHub and Hugging Face, licensed under MIT. While there’s no official API announced yet, a demo is available for users to explore its capabilities.

DeepSeek-R1 Gains Support from Perplexity AI

In another milestone, Perplexity AI has integrated DeepSeek-R1, calling it the “world’s most powerful reasoning model.” CEO Aravind Srinivas confirmed its support alongside OpenAI’s o1 AI model, making it accessible to all Perplexity users.

DeepSeek assures that data security is a top priority—its models are hosted in the US, addressing concerns about data being sent to Chinese servers.

OpenAI CEO Sam Altman Responds to DeepSeek’s Rapid Growth

DeepSeek’s rise in the AI industry has caught the attention of OpenAI CEO Sam Altman, who acknowledged the DeepSeek-R1 model as “impressive,” especially for its cost-effectiveness. He added that OpenAI would soon release more advanced models, embracing the competition.

Did DeepSeek’s AI Claims Impact Nvidia’s Stock?

On Monday, Nvidia’s stock plunged by 13%, erasing $465 billion in market value—its worst single-day decline since going public in 1999. Market analysts speculate that DeepSeek’s claim of developing R1 without expensive GPUs and at a cost of under $6 million may have triggered investor concerns about future AI hardware demand.

Final Thoughts

With DeepSeek Janus Pro 7B, the company is setting new standards in open-source AI development. By offering free and unrestricted access to its cutting-edge models, DeepSeek is rapidly emerging as a serious competitor to industry leaders like OpenAI. As the AI landscape continues to evolve, all eyes are on DeepSeek’s next big move.