Login
iscover
Share this article on:
Powered by
MOMENTUM
MEDIA
Powered by MOMENTUMMEDIA
The Chinese AI giant’s new and improved model could potentially run on a single GPU.
DeepSeek has updated its R1 reasoning AI model, according to a WeChat message earlier this week.
DeepSeek uploaded the updated model, R1-0528, on the Hugging Face developer platform, and while it does not provide a description of the upgrade, it is now under a permissive MIT license, allowing it to be used commercially.
The model is still based on the original V3 architecture, but it has undergone additional training and used increased computational resources.
The DeepSeek announcement calls the update a “minor” version upgrade, which reportedly improves its reasoning and inference capabilities, as well as its ability to deal with complex tasks. This allows it to compete better with OpenAI’s o3 reasoning models and Google’s Gemini 2.5 Pro model.
The upload to Hugging Face means that DeepSeek has stuck to its plan of making its technology open source, maintaining a leg up on the competition at OpenAI.
While media has reported that the new R1 model could potentially run on a single GPU, it is unlikely that consumer-grade hardware would be powerful enough thanks to its 685 billion parameters in size.
DeepSeek has, however, released a smaller variant called R1-0528-Qwen3-8B, that is only 8 billion parameters in size.
According to Yiren Lu, a solutions engineer for Modal, 2GB of GPU memory is needed per 1 billion parameters, meaning a single GPU that has at least 16 GB of VRAM, such as the NVIDIA RTX 3090, RTX 4090, RTX 5060Ti and plenty more,
However, concerns with DeepSeek regarding security and propaganda is still sure to limit its reach.
The Chinese AI has been banned by Microsoft, AusPost, the NBN, the ABC, the Trump administration and many state and national governments.
Despite this, DeepSeek shook up the market majorly upon release. NVIDIA’s stock price fell almost 18 per cent, resulting in a US$600 billion (roughly A$1 trillion) market cap drop, the largest in Wall Street history.
Additionally, former Intel CEO Pat Gelsinger has commended DeepSeek for creating a lower-cost model with performance rivalling OpenAI.
“Wisdom is learning the lessons we thought we already knew. DeepSeek reminds us of three important learnings from computing history,” said Gelsinger.
“1) Computing obeys the gas law. Making it dramatically cheaper will expand the market for it. The markets are getting it wrong, this will make AI much more broadly deployed.
“2) Engineering is about constraints. The Chinese engineers had limited resources, and they had to find creative solutions.
“3) Open wins. DeepSeek will help reset the increasingly closed world of foundational AI model work. Thank you DeepSeek team.”
Gelsinger also said that his start-up, Gloo, is already making use of DeepSeek R1.
“My Gloo engineers are running R1 today,” he said. “They could’ve run o1 – well, they can only access o1, through the APIs.”
With DeepSeek, Gloo said it will rebuild its Kallm AI from scratch “with our own foundational model that’s all open source” in two weeks, rather than paying for and using OpenAI.
Furthermore, when asked if he would be excited with DeepSeek if he was still the CEO of Intel, Gelsinger answered “yes”.
Be the first to hear the latest developments in the cyber industry.
