DeepSeek has announced that it trained its latest AI model, DeepSeek V3, at a cost of $5.58 million over a two-month period. This model boasts 671 billion parameters and was developed using fewer computing resources compared to similar models from larger tech companies like Meta and OpenAI. Despite the reported cost, some experts suggest that this figure may not encompass the total expenses incurred during the model's development, as it excludes prior research, experiments, and testing. The company’s previous model, R1, was released in late December and was also claimed to have been developed at a cost of approximately $5.57 million. Critics have raised concerns that the $5.6 million figure may serve as a marketing strategy, as the overall investment for DeepSeek's models could be higher when considering all developmental stages and infrastructure costs.