DeepSeek launches V3 update with 700GB MIT-licensed model, outperforms OpenAI and rivals in math and code

Following the success of its earlier models, DeepSeek is back with a big update — literally. The Chinese AI startup on Monday rolled out DeepSeek-V3-0324, a hefty 700GB open-source model that’s turning heads in the AI community. Released under the MIT license, the update signals a strong move toward transparency and open collaboration.

Without any formal announcement on X, DeepSeek quietly released DeepSeek-V3-0324 on Hugging Face.

A wild Deepseek has appearedhttps://t.co/9JfI9DqmFz

— Teknium (e/λ) (@Teknium1) March 24, 2025

DeepSeek Releases V3-0324 Update

Currently available on Hugging Face and OpenRouter, DeepSeek-V3-0324 makes it easy for developers and researchers to put it to the test. Reactions across platforms suggest genuine excitement, with many keeping a close eye on what DeepSeek does next — especially its upcoming Reasoner V2 model.

What’s surprising isn’t just the size — it’s how well the model performs. Early users say DeepSeek V3 is holding its own against models from giants like OpenAI, especially in areas like mathematics, coding, and frontend design. Some are even saying it’s outperforming them in key tasks.

When asked about the release, DeepSeek described it as a small update. Developers, though, are seeing something much bigger. For example, one user on X put it bluntly: “New DeepSeek V3 is not just a small update! It’s a world of difference in terms of frontend design! This is literally first try, looks even better than what R1 is capable of.”

New deepseek V3 is not just a small update! It’s a world of difference in terms of frontend design! This is literally first try, looks even better than what R1 is capable of pic.twitter.com/K1QvDTiCYZ

— Phil (@phill__1) March 24, 2025

Meanwhile, this release isn’t just a technical update. It’s part of a broader shift, as more developers look beyond U.S.-based models. DeepSeek’s V3 drop has sparked new conversations about whether China is starting to take the lead in certain areas of AI development. That alone is enough to make the industry sit up and pay attention.

In January, DeepSeek overtook ChatGPT to become the highest-rated free app on Apple’s App Store in the U.S. This achievement came just weeks after its January 10 launch, fueled by the growing popularity of the DeepSeek-R1 AI model.

News of DeepSeek’s success triggered a massive sell-off and sent shockwaves through the U.S. stock market, wiping out a staggering $1 trillion in market capitalization as investors began questioning the sustainability of AI chip spending.

The buzz around DeepSeek grew following reports that its V3 model had outperformed Meta’s Llama 3.1, OpenAI’s GPT-4o, and Alibaba’s Qwen 2.5 on third-party benchmarks—all at a significantly lower cost.