Hangzhou AI company DeepSeek has released its latest large-scale language model V3, which achieves similar performance to OpenAI's GPT-4 and Anthropic's Claude 3.5Sonnet at a low cost ($5.6 million), attracting industry attention. This achievement highlights DeepSeek's strong research and engineering capabilities under limited resources, and also provides new ideas for the research and development of large-scale language models, challenging the traditional concept that high investment can achieve high returns. The success of the V3 model may indicate more possibilities for future development in the AI field.
Recently, DeepSeek, an AI company located in Hangzhou, released its latest large-scale language model-V3. The performance of this open source model in multiple benchmark tests is close to that of OpenAI's 4o and Anthropic's Claude3.5Sonnet, which has impressed the industry. Compared with the hundreds of millions of dollars invested by its American counterparts, the total cost of DeepSeek's V3 model is only US$5.6 million. This huge difference is eye-catching.
Liang Wenfeng, CEO of DeepSeek, said that funding has never been a problem for them. Although V3 is trained based on the H800 chip, the DeepSeek team can still demonstrate strong research and engineering capabilities despite limited resources.
Andrej Karpathy, a pioneer in the field of AI, commented that DeepSeek's investment budget was "what a joke", but the final result was "highly impressive research and engineering under resource constraints."
AGI is regarded as the "Holy Grail" of AI research, capable of surpassing humans in solving problems and performing tasks. Relevant experts predict that once the technology matures, the first country to implement AGI will have huge economic, scientific and security advantages.
The success of the DeepSeek V3 model not only proves that breakthrough progress can be achieved at low cost, but more importantly, it provides a new perspective and direction for global AI research and development, and also brings opportunities for developing countries to catch up with advanced levels in the field of artificial intelligence. gained hope. In the future, there may be more similar "miracles" happening.