DeepSeek V3 Now is the best non-curated AI model on the market, Openai, Grok and Google, AI Benchmarker Manual analysis release Tuesday.
although AI company With the release of the new update, the highest trade blow, which makes the DeepSeek V3’s achievements noteworthy, this is the first time that the open model leads. Human analysis says its intelligent index measures responses in reasoning, knowledge, mathematics, and coding.
Open weighting means that DeepSeek V3 is fully open source and the entire model is publicly available for user adjustment and modification. Openai is unwilling to give up all the secrets of Chatgpt despite its name. AI companies keep training closely on the infrastructure of their training models and how they operate. have Advantages of open and closed methods For software development, the choice between the two is usually determined by a business interest.
DeepSeek leads: DeepSeek V3-0324 is now the most scoring non-dispute model
This is the first time that the open weight model is a leading non-controversial model, and it is a milestone in open source.
DeepSeek v3-0324 has jumped forward by 7 points in manual analysis… pic.twitter.com/t7gejnqibs– Manual Analysis (@Artaveranlys) March 25, 2025
Given that V3 is not a “reasonable” AI model, it is not powerful. At the same time, it’s faster, which means running in most applications is cheaper and more feasible.
The DeepSeek V3 was launched last December, with its latest update increasing competition among U.S. AI companies. DeepSeek gained a standout earlier this year When it releases its R1 inference model for free. This is the first time a high-level inference model that recursively returns and checks the answer before giving the final output, which is widely used by the masses. The operating costs of the DeepSeek R1 are also impressive, which shows that the company has made innovations in efficiency.
However, the world of inference AI models is more suitable for research or tasks that require large data sets. The non-resonant models of GPT-4.5 and Google Gemini 2.0 (such as GPT-4.5 and Google Gemini 2.0) are more suitable for most applications because they are faster and cheaper. The rise of DeepSeek V3 shows that Chinese artificial intelligence companies can be competent for American companies while also retaining open source code.
This threat is why Openai is pushing the Trump administration to raise restrictions on training through copyrighted materialsbelieves that if there are still fair usage restrictions, American AI companies will not be able to compete with China.
