There was no press conference, no overwhelming publicity,On March 24, DeepSeek V3-0324 version was quietly launched.

https://huggingface.co/deepseek-ai/DeepSeek-V3-0324
Although this update does not belong to the new generation of R2, its improvement is far beyond expectations, especially in terms of programming ability, front-end UI design and context understanding, it is already comparable to Claude 3.7 Sonnet.
Another important upgrade is that DeepSeek continues to adhere to the spirit of open source and directly adopts the MIT open source protocol, allowing developers to freely modify and commercialize,completely breaking the closed-source AI monopoly on top code models.
What does all this mean?
What has evolved in DeepSeek V3-0324?
DeepSeek V3 itself is recognized by the industry as a cost-effective open source model. Although the 0324 version is only a "minor version update", its improvement is not small. Yes, DeepSeek is so low-key.
Parameter scale has been further improved: the model parameters have increased from 671 billion to 685 billion. Although the increase is not large, the optimized computing efficiency has brought smoother reasoning capabilities.
Code capabilities have been greatly improved: The new version of DeepSeek V3 has significantly improved its capabilities in front-end development, website construction, and UI design. Everyone basically believes that it is comparable to Claude 3.7 Sonnet.
Stronger mathematical reasoning: Improved multi-round dialogue and logical reasoning capabilities make it more stable in solving mathematical problems, and some problem-solving capabilities are close to the reasoning model.
Open source license is more free: The adoption of the MIT license allows developers to modify, optimize, and commercialize at will, greatly lowering the threshold for AI applications.
Actual performance: Is the new version V3 really that powerful?
In the test of professional users, DeepSeek V3-0324's code capabilities have reached the world's strongest open source AI level, and even surpassed Claude 3.7 in some scenarios.
Some netizens said that in the Aider benchmark test, DeepSeek V3-0324 has improved significantly and is close to Claude Sonnet 3.7.

Some netizens also released the test results of the front-end design, and it can be seen that the design capability of the new version V3 even exceeds R1. 
The following example is more obvious, directly comparing several top models. This netizen asked them to generate the code for animated weather cards.



The results are very intuitive. The new version V3 is far superior to the original V3 and R1, and is directly comparable to Claude 3.7.
An AI developer lamented: "Since June 2024, no LLM on the market has been able to get close to Claude 3.7 in front-end code generation, but DeepSeek V3-0324 finally did it."
Open source killer: MIT license, AI entering the free era?
The biggest "hidden easter egg" of this update is that DeepSeek completely unlocked the MIT license, making V3-0324 a completely open, commercially available top AI model.
Why is the MIT license so important?
No barriers to commercial use: Enterprises and individual developers can use it freely without paying expensive licensing fees.
Free model transformation: You can modify and fine-tune it at will, and even use it in your own products or projects without official restrictions.
Promote the development of AI ecology: Compared with the closed-source business model of OpenAI's GPT-4.5 and Claude 3.7, DeepSeek allows more entrepreneurial teams and individual developers to enjoy top AI capabilities equally, truly realizing "technical equality".
Some foreign netizens said: "DeepSeek is doing what Meta promised but did not do, making AI really open source and free."
But some are happy and some are worried. If open source AI like DeepSeek continues to soar, what else can closed source AI rely on to attract users?
AI competition is heating up, and business models are "trembling"
The release of DeepSeek V3-0324 not only enhances the status of open source AI in programming and mathematics, but also brings unprecedented pressure to closed source AI companies such as OpenAI and Anthropic.
Previously, closed source AI could attract high-end users with "stronger capabilities", but now, the free and open source DeepSeek V3 is close to Claude 3.7, and even better in some areas.
Some overseas users said frankly: "Seeing the progress of these open source models, I really hope that commercial AI companies can lower their prices."
The upgrade of DeepSeek has also made people begin to re-examine the future of the AI industry - When high-performance AI gradually moves towards open source, does the business model of AI need to be completely changed?
With the evolution of V3, will R2 be far behind?
Then, the next question is - Will R2 be far behind?
Many DeepSeek users said thatif V3-0324 is already so powerful, then R2 may really overturn the entire AI ecosystem.
In the coming months, the field of AI may usher in greater changes, and DeepSeek has already stood at the center of the storm.