DeepSeek Launches V4 Pro as China’s AI Race Intensifies
Chinese artificial intelligence company DeepSeek has officially released DeepSeek V4 Pro, moving its flagship model out of preview as competition among China’s leading AI developers accelerates.
The new version, DeepSeek-V4-Pro-0813, became generally available on August 13 across DeepSeek’s app, website and API. The company says the update delivers major improvements in agent-based tasks, particularly coding, tool use and complex workflows.
DeepSeek V4 Pro supports a one-million-token context window and offers both thinking and non-thinking modes. Its API also supports tool calls, JSON output and the Responses API format, allowing developers to integrate the model into agent and coding workflows.
The company has also added three levels of reasoning effort — low, high and max — allowing users to adjust how much computation the model applies depending on the complexity of a task.
DeepSeek published a series of benchmark results showing substantial gains over its earlier V4 Pro preview, including stronger performance across coding-agent and tool-use tests. Those figures are DeepSeek’s own benchmark results and should not be treated as independent verification of performance.
The launch is particularly notable because DeepSeek’s smaller and cheaper V4 Flash model had unexpectedly beaten the V4 Pro preview in several independent evaluations following its July release. Reuters reports that the new Pro version is being closely watched to see whether DeepSeek has closed that gap.
DeepSeek is also changing its pricing strategy.
From August 16 at 16:00 UTC, the company will introduce separate peak and off-peak API pricing, with off-peak rates set at half the peak price. For V4 Pro, one million uncached input tokens will cost $1.32 during peak periods and $0.66 off-peak, while one million output tokens will cost $3.96 at peak and $1.98 off-peak.
