DeepSeek V4 Pro is officially online

On 13 August, DeepSeek officially went live on V4 Pro, an official version updated to DeepSeek-V4-Pro-0813. The model supports the 1M Token context, 384K Token maximum output and defaults to start thinking mode, as well as JSON output, tool call, Responsions API and Anthropic API compatibility, which can be used for long document, code repository and Agent task processing。