DeepSeek announced the release of DeepSeek-V4-Pro, now available on DeepSeek Chat under "Expert Mode" and via API. The model includes major agent upgrades, flexible reasoning effort, and native support for OpenAI's Responses API, optimized for Codex.
SynthePulse Insight · AI deep reading
Version 1 · 1 source
DeepSeek released V4-Pro on August 13, 2026, entering the market with 'Expert Mode' and flexible reasoning investment, but its naming and pricing strategy has sparked discussion about product positioning and maturity.
On August 13, 2026, DeepSeek officially announced the launch of DeepSeek-V4-Pro, immediately available under 'Expert Mode' in DeepSeek Chat, with API access also opened. This release was reposted by third-party account TestingCatalog, with original information from DeepSeek's official account.
The introduction of 'Expert Mode' has sparked speculation about the reasons behind it. One user commented that this tiered approach might stem from capacity limitations, or because the model is not yet ready as a default option, rather than a direct replacement for the base model. This suggests a cautious product strategy.
DeepSeek's official announcement highlighted the 'flexible reasoning investment' feature of V4-Pro and V4-Flash, offering low, high, and maximum settings, suitable for simple tasks, daily agent workflows, and complex tasks, respectively. This design aims to let users adjust computational resources based on task requirements.
Additionally, V4-Pro natively supports the OpenAI Responses API and is optimized for Codex, providing one-click setup. This integration may aim to attract existing OpenAI users and reduce migration costs.
Following the release, some users reported switching to the V4-Pro endpoint, noting its output pricing of $0.87 per million, which they believe makes automated nightly test generation economically viable. This price point could be a key factor for adoption by small and medium teams.
However, some comments also questioned the naming strategy, suggesting that launching the V4-Pro product tier before extensive benchmarking of V3 reflects more of a marketing strategy than the model's actual capabilities. This viewpoint implies that product naming may be ahead of technical validation.
This report is based on an announcement from DeepSeek's official X account, reposted by third-party account TestingCatalog, and includes user comments. The official announcement confirmed the model release, feature set, and API support, but did not provide benchmark data or performance details. Pricing and opinions in user comments are personal claims and have not been independently verified.
The release of DeepSeek-V4-Pro marks its advancement in agent workflows and API integration, but the tiering of 'Expert Mode' and naming strategy suggest caution and market considerations in productization. Pricing and feature details may attract developers, but the model's true performance still requires independent benchmark testing to verify.
Primary report
Primary source