Amid the rapid iteration in the field of artificial intelligence, DeepSeek has once again become a focal point in the industry. In addition to quietly releasing the V4.1Flash, which features a significant architectural overhaul and is essentially a generation update, industry rumors suggest that it will soon launch its highly anticipated new large model - DeepSeek Code2.0.

Looking back at recent actions, DeepSeek officially launched the V4.1Flash this week. Although the official version number was not changed, the upgrade in its underlying architecture actually reached the level of V5. At the same time, the official has also confirmed that it will subsequently release the V4.1Pro series, although specific details about the upgrades have not been fully disclosed. Considering that the performance of the previous V4Pro-0813 final version did not meet expectations, the market is maintaining a rational expectation for the actual performance of this series.

Outside of general-purpose large models, leaks indicate that DeepSeek will introduce a new major product, DeepSeek Code2.0, in September. The most remarkable feature of this flagship product is its parameter count exceeding 3 trillion. From an industry perspective, the current products with the largest parameter counts include K3's 2.8 trillion, Qwen3.8Max's 2.4 trillion, and DeepSeek V4Pro's 1.6 trillion. Considering that V4.1Flash achieved a doubling of parameters through its new structure, it makes logical sense for Code2.0 to jump from 1.6 trillion to 3 trillion.

In terms of core performance and functional positioning, DeepSeek Code2.0 is expected to maintain the tradition of open-source and openness, with its design focus shifting entirely to "computer control," meaning that AI will automatically operate the computer on behalf of users to complete various complex tasks. Industry rumors suggest that its performance metrics may be able to outperform the current strongest frontier-level large models Mythos5.1 and GPT-6Astra.

Although the external world has high expectations for this new generation of flagship large model focused on code and logic, its ultimate performance still needs to be tested by the market. As the V4.1Flash, followed by the Pro and Code2.0 series, gradually emerge, DeepSeek is continuously challenging the top line of global AI through its multi-route product strategy.