DeepSeek’s launch and unveiling of the DeepSeek-V3 on Monday made tech firms like Nvidia and Oracle witness steep drops after the China-based company revealed that its model was the best open-source model.
DeepSeek, early on January 28, Lunar New Year’s Eve, released Janus-Pro and JanusFlow, two new multimodal frameworks.
Janus-Pro is an improved version of Janus, intended to serve as a single foundation for multimodal creation and interpretation. The concept enhances performance and adaptability across a range of jobs by decoupling visual encoding.
In benchmark testing, it has done better than OpenAI’s DALL-E 3 image-generation model. Janus-Pro is open-source, as is the case with earlier Janus models. There are two versions of the model: JanusPro 7B, which has 7 billion parameters, and JanusPro 1.5B, which has 1.5 billion parameters.
Accusations Over Data Usage
OpenAI, the company that created ChatGPT, has expressed dissatisfaction over competitors, notably those in China, rapidly advancing their own artificial intelligence (AI) capabilities by leveraging their work.
This week, DeepSeek, a Chinese AI model that can mimic ChatGPT’s performance, and reportedly built at a fraction of ChatGPT’s price, appeared out of nowhere. It severely undermined OpenAI’s and other US companies’ position as global leaders in AI.
According to a report from Bloomberg, Microsoft, a significant investor in OpenAI, is looking into whether its data has been used in an unauthorized manner.
The Bottom Line
The tale of DeepSeek is not just about getting more done with less; it’s also about demonstrating that hard work, innovation, and attention can outperform even the most extravagant budgets. One clever move at a time, DeepSeek is out there playing 4D chess while billion-dollar corporations are busy splurging on issues.
Point them to DeepSeek the next time someone tells you that you need a lot of money to compete in the AI market. Because sometimes all it takes is 5.6 million dollars and a great deal of creativity.
