Tan Dai from Volcano Engine: Technology-driven large model price reduction, the industry should focus on fundamentals like DeepSeek

On February 13th, Jinshi Data reported that in response to the viewpoint of malicious price wars among large models in the market, Tan Dai, president of Volcano Engine, stated through his social media that the price reduction of large models is achieved through technological innovation, and everyone should focus on fundamental skills like DeepSeek and refrain from baseless speculation. Tan Dai mentioned that the pre-training cost and inference cost of the Doubao 1.5Pro model are lower than those of DeepSeek V3, and significantly lower than other domestic models, resulting in a very good gross profit at the current price. "Manufacturers both domestically and internationally are relying on technological innovation to reduce model prices. We have only achieved the price level of Gemini 2.0 Flash, and this price is entirely dependent on technological advancements." According to data, the input price for Doubao-1.5-pro-32k is 0.8 yuan per million tokens, with a cache hit price of 0.16 yuan and an output price of 2 yuan