The 'Models & Pricing' page of the DeepSeek official website API documentation has added the DeepSeek-V4-Pro-0813 model version. As a flagship product, it features a context window of 1 million tokens and a maximum output of 384,000 tokens. It enables the thinking mode by default and provides a non-thinking mode endpoint to handle latency-sensitive calls. Currently, the official has not yet released its changelog.