DeepSeek’s groundbreaking AI model V3 outperforms Meta and OpenAI, achieving unparalleled performance with fewer resources

Chinese AI startup DeepSeek made headlines with the release of DeepSeek V3, an advanced large-scale language model. Boasting 671 billion parameters, the model outperformed prominent AI models such as Meta’s Llama 3.1 and OpenAI’s GPT-4o in benchmark tests that assess text understanding, coding, and problem solving. This result is a major step for China’s AI industry.

The Hangzhou-based company revealed in a post on WeChat that DeepSeek V3 utilized only 2.78 million GPU hours and was developed at a staggering cost of just $5.58 million. In comparison, Meta’s Llama 3.1 required 30.8 million GPU hours. DeepSeek relied on Nvidia’s H800 GPUs, which are tailored for the Chinese market, circumventing U.S. sanctions that block access to advanced chips.

Also read: After 6th generation stealth fighter jet, China unveils world’s largest amphibious ship in big challenge to US; Here are the specifications and all details

Computer scientist Andrei Karpathy praised the achievement on X (formerly Twitter), noting that DeepSeek was able to create frontier-grade models with minimal resources. According to DeepSeek’s technical report, the V3 model not only outperformed Meta and Alibaba’s models, but also produced results comparable to OpenAI’s GPT-4o and Amazon-backed Anthropic’s Claude 3.5 Sonnet.

DeepSeek will be spun off from High-Flyer Quant in 2022 with a focus on cost-effective AI development. The company’s Fire Flyer GPU clusters have helped drive innovation. DeepSeek aims to democratize AI, offering a model for third-party application development alongside chatbot services.

Also read: Apple withdraws iPhone 14 and SE from Europe as USB-C mandate changes things

FAQ:

What is DeepSeek V3?
DeepSeek V3 is a large-scale language model (LLM) developed by Chinese startup DeepSeek. DeepSeek V3 has 671 billion parameters and is designed to understand and generate text more effectively than many existing models. How does DeepSeek V3 compare to other AI models?
DeepSeek V3 outperforms models such as Meta’s Llama 3.1 and OpenAI’s GPT-4o on a variety of benchmark tests including text generation, coding, and problem solving. This rivals the capabilities of some of the most advanced AI systems available.