DeepSeek has already taken the internet by storm when reports began circulating claiming that it outperformed the world's most advanced AI models. Notably, DeepSeek spent less than $6 million to train its AI models, compared to the $100 million that OpenAI invested to do the same.
This resulted not only in DeepSeek to become the number one free app in the App Store, but caused and a historic “disappearance” of over $400 billion in NVIDIA’s US market.
Soon after, the AI chatbot gained a huge amount of traffic that caused DeepSeek's server outages and performance issues.
Despite the challenges, the Chinese AI lab has forged ahead, unveiling a groundbreaking open-source AI model called Janus-Pro. The new image generation model reportedly outperforms OpenAI's DALL-E, Stability AI's Stable Diffusion, and other image generation models in several benchmarks.
Janus-Pro is an update to Janus, which was released late last year.
Janus-Pro comes in a variety of sizes, from a compact 1 billion-parameter version to a 7 billion-parameter version, close to the size of a 3,5L SD. The largest model, the Janus-Pro-7B, excels in both image generation and analysis, outperforming leading competitors PixArt-alpha, Emu3-Gen and SDXL in the industry benchmarks GenEval and DPG-Bench, according to information shared by DeepSeek.
The Janus-Pro-7B model is available for free download by Huggingface, a popular place for artificial intelligence and machine learning.
George is still wondering what he is doing here….
