Recently, the AI company DeepSeek, based in Hangzhou, launched its latest large language model — V3. This open-source model has shown performance in several benchmark tests that is comparable to OpenAI's 4o and Anthropic's Claude 3.5 Sonnet, which has caught the industry's attention. In contrast to the hundreds of millions of dollars invested by their American counterparts, the total cost of DeepSeek's V3 model is only $5.6 million, highlighting a significant difference.
Image Source Note: The image is generated by AI, and the image is licensed by the service provider Midjourney.
DeepSeek's CEO Liang Wenfeng stated that funding has never been an issue for them. Although V3 is trained on H800 chips, DeepSeek's team has still demonstrated strong research and engineering capabilities despite limited resources.
AI pioneer Andrej Karpathy commented that DeepSeek's investment budget is "really a joke," yet the final results are "highly impressive research and engineering under resource constraints."
AGI is considered the "holy grail" of AI research, capable of surpassing humans in problem-solving and task execution. Experts predict that once the technology matures, the first country to achieve AGI will hold significant advantages in economics, science, and security.
