Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn MoreChinese AI startup DeepSeek has quietly released a new large language model that’s already sending ripples through the artificial intelligence industry — not just for its capabilities, but for how it’s being deployed. The 641-gigabyte model, dubbed DeepSeek-V3-0324, appeared on AI repository Hugging Face today with virtually no announcement, continuing the company’s pattern of low-key but impactful releases.What makes this launch particularly notable is the model’s MIT license — making it freely available for commercial use — and early reports that it can run directly on consumer-grade hardware, specifically Apple’s Mac Studio with M3 Ultra chip.The new Deep Seek V3 0324 in 4-bit runs at > 20 toks/sec on a 512GB M3 Ultra with mlx-lm! pic.twitter.com/wFVrFCxGS6 — Awni Hannun (@awnihannun) March 24, 2025“The new DeepSeek-V3-0324 in 4-bit runs at > 20 tokens/second on a 512GB M3 Ultra with mlx-lm!” wrote AI researcher Awni Hannun on social media. While the $9,499 Mac Studio might stretch the definition of “consumer hardware,” the ability to run such a massive model locally is a major departure from the data center requirements typically associated with state-of-the-art AI.DeepSeek’s stealth launch strategy disrupts AI market expectationsThe 685-billion-parameter model arrived with no accompanying whitepaper, blog post, or marketing push — just an empty README file and the model weights themselves. This approach contrasts sharply with the carefully orchestrated product launches typical of Western AI companies, where months of hype often precede actual releases.Early testers report significant improvements over the previous version. AI researcher Xeophon proclaimed in a post on X.com: “Tested the new DeepSeek V3 on my internal bench and it has a huge jump in all metrics on all tests

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More. Last week, Chinese startup DeepSeek sent shockwaves in the AI community with its frugal yet highly performant open-source release, DeepSeek-R1. The model uses pure reinforcement learning (RL) to match OpenAI’s o1 on a range of benchmarks, challenging the longstanding notion that only large-scale training with powerful chips can lead to high-performing AI. However, with the blockbuster release, many have also started pondering the implications of the Chinese model, including the possibility of DeepSeek transmitting personal user data to China. The concerns started with the company’s privacy policy. Soon, the issue snowballed, with OpenAI technical staff member Steven Heidel indirectly suggesting that Americans love to “give away their data” to the Chinese Communist Party to get free stuff.The allegations are significant from a security standpoint, but the fact is that DeepSeek can only store data on Chinese servers when the models are used through the company’s own ChatGPT-like service. If the open-source model is hosted locally or orchestrated via GPUs in the U.S., the data does not go to China. Concerns about DeepSeek’s privacy policyIn its privacy policy, which was also unavailable for a couple of hours, DeepSeek notes that the company collects information in different ways, including when users sign up for its services or use them. This means everything from account setup information — names, emails, numbers and passwords — to usage data such as text or audio input prompts, uploaded files, feedback and broader chat history goes to the company.But, that’s not all