Chinese startup DeepSeek has taken the tech world by storm with the launch of its innovative AI model, resulting in a significant decline in the stock prices of American tech companies. Among the hardest hit is NVIDIA, a leading AI chip maker. The primary reason behind this downturn is that DeepSeek's AI model comes at a lower cost compared to other popular platforms like ChatGPT, Google Gemini, and Meta AI. This affordability has made DeepSeek's model a hot topic globally.
DeepSeek has introduced two AI models, named R1 and R1 Zero. These models are built on open-source licenses and are available for free to users. Notably, DeepSeek’s AI model stands apart from generative AI counterparts like ChatGPT and Google Gemini. Its impact is felt not just by American firms but also by Chinese giants Alibaba and Baidu, which have invested billions into their own AI developments.
What sets DeepSeek R1 apart?
Launched just a month ago, DeepSeek R1 is a reasoning model that boasts a price point lower than its competitors. This advanced language-based AI model leverages augmented reasoning and analytical capabilities, built on hybrid architecture similar to V3. The cost of using DeepSeek R1 is remarkably low, at USD 0.55 (approximately Rs 47) for every million input tokens and USD 2.19 (around Rs 189) for each million output tokens.
The rapid development of this Chinese AI model has also garnered attention, taking just two months to create. This contrasts sharply with companies like Google, Microsoft, and Meta, which have spent over six years and invested heavily in their AI initiatives. Microsoft CEO Satya Nadella has weighed in on DeepSeek's model, stating it should be taken seriously, while OpenAI CEO Sam Altman praised the Chinese innovation on his X handle, describing it as quite impressive.
Unlike AI tools like ChatGPT and Google Gemini that operate on Large Language Models (LLM) and are generative in nature, DeepSeek's AI model focuses on reasoning. While the former provide responses based on existing internet data they were trained on, DeepSeek’s model is capable of responding to commands with cognitive reasoning. This unique approach could pose a significant challenge to other AI platforms in the future.
ALSO READ: Google Pixel 9a price: Higher storage options likely to cost more than last year's model