SiliconFlow Faces High Demand for DeepSeek-V2.5 as V2-Fast Price Drops to 0.6 RMB
Chinese AI platform SiliconFlow (Seedance) has experienced a surge in traffic following the launch of the DeepSeek-V2.5 model. To remain competitive, the platform has also slashed the API price of the DeepSeek-V2-Fast model to just 0.6 RMB per million tokens. This move intensifies the ongoing price war in the Chinese large language model (LLM) market, significantly lowering the barrier to entry for developers. By offering highly capable models at extremely low costs, platforms like SiliconFlow are accelerating the commercial integration of AI. DeepSeek-V2.5 is an upgraded version that merges the chat and coding capabilities of DeepSeek-V2-Chat and DeepSeek-Coder-V2. The price reduction to 0.6 RMB per million tokens makes the older V2-Fast model one of the most cost-effective options available for high-throughput tasks.
## BACKGROUND
SiliconFlow is an AI infrastructure company that provides developers with optimized APIs to deploy and run over 200 large language and multimodal models. DeepSeek is a Chinese AI research company funded by High-Flyer that focuses on developing highly efficient, open-weight LLMs, which have disrupted the industry due to their low training costs and high performance.