Open-source AI is shaking up the dominance of big tech and the rise of Chinese startup DeepSeek is a good example to justify the statement. The stock market on Monday witnessed a sharp decline and Nasdaq dropped nearly 3%. The rapid rise of DeepSeek is the reason for such a sudden dip as the new AI Chatbot has gained worldwide attention soon after its launching of two modesl -DeepSeek-V3 and DeepSeek-R1. The fact is interesting that DeepSeek has achieved the popularity at just a fraction of the investment compared to other AI models supported by giants like OpenAI, Meta and Google.<\/p>\n
DeepSeek was founded by entrepreneur Liang Wenfeng and he is basically known for running a hedge fund High Flyer in China. The startup has managed to make a name for itself in AI research even after operating with significantly lower funding than its American counterparts. This has gained widespread acceptance as developers and businesses now have access to high-quality AI tools without being locked into big tech ecosystems.<\/p>\n
The standout feature of DeepSeek AI models is their efficiency. DeepSeek-V3 has been trained on a budget of just $5 million and has outperformed leading models like GPT-4o and Claude 3.5 Sonnet in multiple benchmarks. One key reason is its unique architecture called Mixture-of-Experts (MOE). It does not rely on a single massive model, but MOE splits tasks among specialized smaller models. Hence, the system is comparatively more efficient and more cost-effective. DeepSeek-V3 has also been trained on an extensive dataset of 14.8 trillion tokens to develop a deep understanding of language and problem-solving tasks.<\/p>\n
DeepSeek-R1 is capable of real-time reasoning and the deep thinking operates behind closed doors. It is open-source to help developers study and improve upon it freely. R1 has either matched or exceeded OpenAI\u2019s performance in tasks such as math, coding and general knowledge. It also costs 90-95% less to run than OpenAI\u2019s equivalent models.<\/p>\n
Another factor for the success of DeepSeek is its low costs hardware choices. Companies like OpenAI rely on the most advanced AI chips like NVIDIA\u2019s H100, but DeepSeek has made it possible with a less powerful version\u2014NVIDIA\u2019s H800. The chips were created after the U.S. government imposed restrictions on the export of high-end AI chips to China due to national security concerns. However, the engineers at DeepSeek worked around such limitations by optimizing their software to make the most of the hardware available to them. They used techniques like Multi-Head Latent Attention (MLA) and Auxiliary-Loss-Free Load Balancing to maintain high performance without requiring expensive computing resources.<\/p>\n
DeepSeek\u2019s success is a major challenge to the AI industry\u2019s status quo. While companies like OpenAI, Google, and Meta have been spending billions to stay ahead, DeepSeek has proven that innovation does not always require massive budgets. By making its models open-source, it has also forced a discussion on whether AI should remain locked behind corporate paywalls or be accessible to everyone. As the debate around open-source AI versus proprietary AI continues, DeepSeek\u2019s rise shows that big tech\u2019s monopoly on AI may not be as secure as it once seemed.<\/p>\n","protected":false},"excerpt":{"rendered":"