Deepseek stuns world with top performing model on a tiny budget

Deepseek has stunned the world by creating a frontier model on a tiny budget and much simpler compute platform, quickly becoming a major player in the artificial intelligence field, challenging industry giants like OpenAI and GoogleDealroom has a profile for this one. Try Dealroom →. The company's innovative approach to AI development has led to significant advancements in artificial general intelligence (AGI).



Sanctions on China seem to have had the effect of making entrepreneurs more resourceful. DeepSeek's ability to create a frontier model on a tiny budget of US$6 million, using significantly fewer computing resources than its rivals, exemplifies this resourcefulness. This development suggests that external pressures have spurred innovation and efficiency in China's AI sector.



Liang WenfengDealroom has a profile for this one. Try Dealroom →, DeepSeek's founder and CEO, brings a unique perspective to AI development, drawing from his background in quantitative finance and experience at High-Flyer Capital ManagementDealroom has a profile for this one. Try Dealroom →. His leadership style emphasizes curiosity-driven research and a bottom-up approach to innovation.



In November 2024, DeepSeek unveiled DeepSeek-R1, a reasoning AI model that rivals OpenAI's o1. This model employs a "thinking" process to consider queries and fact-check itself, helping it avoid common pitfalls. DeepSeek-R1 performs well on benchmarks like AIME and MATH but does have limitations, including challenges with certain logic problems.



In December 2024, DeepSeek released DeepSeek-V3, an even more advanced model with 671 billion parameters that uses a Mixture of Experts (MoE) architecture. It was trained on 14.8 trillion tokens and generates text three times faster than its predecessor while outperforming many open-source and proprietary models.



DeepSeek's approach is characterized by open-source development, cost-effective solutions, advanced reasoning capabilities, and a commitment to pushing the boundaries of AI technology. This dedication positions DeepSeek as a significant player in the AI landscape, potentially accelerating advancements in the field.



Source: Techcrunch, Gulf News,Paul Covert



NM