What happened
DeepSeek has not released or announced a V5 model, according to the company's official API changelog and news page. The latest available model is DeepSeek-V4.1-Flash-1017, released on October 17, 2026. While rumors of a V5 release have circulated, the company's public roadmap identifies the unreleased V4.1-Pro as the next model in its current architecture family. Reports from Chinese financial outlet Caiwen on November 3, 2026, indicate that V5 training is still in the planning phase, with a potential release expected in 2027.
DeepSeek has not released, previewed, or officially announced a V5 model. The company's API changelog, which serves as the definitive record of its model deployments, concludes with the October 17, 2026, update for DeepSeek-V4.1-Flash-1017. There are no V5 repositories on the company's official Hugging Face account.
The company's public communication identifies V4.1-Pro as the next model in its current architecture family. While DeepSeek previously planned to retire the older V4-Pro, it reversed this decision due to user demand, meaning both V4-Pro-0813 and V4.1-Flash remain active on the API.
Reports from the Chinese financial outlet Caiwen on November 3, 2026, citing industry sources, suggest that V5 is currently in the planning phase. These sources project a release in the first or second half of 2027, contradicting earlier analyst forecasts from Bank of America that had suggested a late 2026 debut.
References to 'deepseek-v5' found in the company's open-source Harness code are confirmed to be test placeholders. Analysis of the code shows these entries are used to test how the application handles unknown model names, with some tests assigning the placeholder a non-functional 9-token .
Why it matters
The distinction between verified product releases and speculative industry rumors is critical for developers and enterprises relying on DeepSeek's infrastructure. By clarifying that V5 is not currently in production, this report prevents misallocation of resources toward non-existent tools. Furthermore, the reported shift toward on Huawei Ascend chips—if realized—would represent a significant milestone for domestic Chinese AI hardware independence. Understanding that current 'leaks' are merely test placeholders in the company's open-source Harness agent allows users to focus on the performance capabilities of the currently available V4.1-Flash model.
The clarification of the V5 status is essential for users who might otherwise rely on speculative timelines. By distinguishing between the company's official roadmap—which prioritizes V4.1-Pro—and viral social media claims, users can make informed decisions about their current AI stack.
The reported plan to use Huawei Ascend 950 chips for is a significant development for the AI industry. If successful, it would mark a shift from using domestic chips primarily for inference to utilizing them for the more computationally intensive pretraining phase, potentially reducing reliance on restricted Nvidia hardware.
DeepSeek's current model, V4.1-Flash, is a 552-billion-parameter mixture-of-experts model that the company claims outperforms its predecessor in cost, speed, and runtime. Understanding that this is the current state-of-the-art for the provider ensures that developers are utilizing the most efficient tools currently available.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
crm_get_transaction(id='4092').Which component of an AI application is the machine-learning model itself?
What to watch next
Observers should monitor DeepSeek’s official API changelog, Hugging Face repository, and the @deepseek_ai X account for the first verified mention of V5. Additionally, the release of the V4.1-Pro model remains the next confirmed step in the company's product cycle. Future developments regarding the use of Huawei Ascend 950 chips for will serve as a key indicator of the company's ability to scale large-model development using domestic hardware, a move that remains unconfirmed by both DeepSeek and Huawei.
The primary indicator for a new model release will be an entry in the official DeepSeek API changelog. Any information not originating from this source or the company's official social media channels should be treated as speculative.
The eventual release of V4.1-Pro will provide insight into the company's scaling strategy for its current architecture family. This release is expected to precede any V5 generation models.
Industry analysts will be watching for confirmation regarding the use of Huawei Ascend chips for . While reports suggest this is the intended path for V5, neither DeepSeek nor Huawei has provided official confirmation of such a partnership for training purposes.