DeepSeek has officially announced the release of DeepSeek-V4-Pro, featuring major agent upgrades, flexible reasoning effort, and native support for OpenAI's Responses API, optimized for Codex. The model is now available on app, web, and via API.
SynthePulse Insight · AI deep reading
Version 1 · 1 source
DeepSeek officially released V4 Pro and V4 Flash on August 13, 2026, confirming that previously leaked benchmarks were accurate. The new models feature enhanced agent capabilities and flexible reasoning effort, with native support for the OpenAI Responses API, but price increases have sparked dissatisfaction among some users.
On August 13, 2026, DeepSeek officially announced the release of DeepSeek-V4-Pro via the X platform. According to third-party reposter Andrew Curran, the official announcement confirmed that previously leaked benchmarks were accurate. This confirmation elevates the leaked data from rumor to official endorsement, but specific benchmark figures were not disclosed in the announcement.
Notably, the official announcement itself did not directly mention the leaked benchmarks; rather, it was reposter Andrew Curran who summarized this in his tweet. Therefore, the claim that 'the leaked benchmarks are accurate' should currently be regarded as a third-party paraphrase, not an official direct statement.
The official announcement listed three major upgrades: first, significantly enhanced agent capabilities, bringing notable productivity gains; second, both V4 Pro and V4 Flash support flexible reasoning effort with three tiers—low, high, and maximum—suitable for simple tasks, everyday agent workflows, and complex tasks, respectively; third, native support for the OpenAI Responses API, optimized for that API.
These upgrades indicate that DeepSeek is actively positioning itself in agent application scenarios, balancing performance and cost through adjustable reasoning effort, while leveraging the OpenAI-compatible API to lower the barrier for developer migration.
In the comments section of the official announcement, user reactions were polarized. One user (Putin is a Virus) expressed disappointment, stating that the benchmarks were poor and the price had increased significantly, and noted that while Flash is decent, at the new price it offers less value than Luna. Another user (Chimpy) described DeepSeek's entry with 'the big whale is coming,' implying its impact.
These comments reflect the market's sensitivity to DeepSeek's pricing strategy and the presence of competitors (such as Luna). However, specific price increases and benchmark figures were not provided in the announcement, so the negative user evaluations cannot be independently verified.
This report is based on DeepSeek's official X announcement and the tweet from third-party reposter Andrew Curran. The official announcement content is first-hand information, but the accuracy of the leaked benchmarks is only relayed by the reposter and has not been directly confirmed by the official source. User comments are personal opinions and do not constitute facts. All specific figures (such as prices and benchmark scores) were not disclosed in the sources.
The official release of DeepSeek V4 Pro confirms the accuracy of leaked benchmarks and brings upgrades in agent capabilities and reasoning effort, but the price increase has sparked dissatisfaction among some users. For developers focused on value, it is advisable to wait for official detailed pricing and benchmark data before making an assessment.
Primary report
Primary source