Large Language Models No Longer Require Powerful Servers
Scientists from Yandex, HSE University, MIT, KAUST, and ISTA have made a breakthrough in optimising LLMs. Yandex Research, in collaboration with leading science and technology universities, has developed a method for rapidly compressing large language models (LLMs) without compromising quality. Now, a smartphone or laptop is enough to work with LLMs—there's no need for expensive servers or high-powered GPUs.
This method enables faster testing and more efficient implementation of new neural network-based solutions, reducing both development time and costs. As a result, LLMs are more accessible not only to large corporations, but also to smaller companies, non-profit laboratories and institutes, as well as individual developers and researchers.
Previously, running a language model on a smartphone or laptop required quantising on an expensive server—a process that could take anywhere from a few hours to several weeks. Quantisation can now be performed directly on a smartphone or laptop in just a few minutes.
Challenges in implementing LLMs
The main obstacle to using LLMs is that they require considerable computational power. This applies to open-source models as well. For example, the popular DeepSeek-R1 is too large to run even on high-end servers built for AI and machine learning workloads, meaning that very few companies can effectively use LLMs, even if the model itself is publicly available.
The new method reduces the model's size while maintaining its quality, making it possible to run on more accessible devices. This method allows even larger models, such as DeepSeek-R1 with 671 billion parameters and Llama 4 Maverick with 400 billion parameters, to be compressed, which until now could only be quantised using basic methods and resulted in significant quality loss.
The new quantisation method opens up more opportunities to use LLMs across various fields, particularly in resource-limited sectors such as education and the social sphere. Startups and independent developers can now implement compressed models to create innovative products and services without the need for costly hardware investments. Yandex is already applying the new method for prototyping—creating working versions of products and quickly validating ideas. Testing compressed models takes less time than testing the original versions.
Key details of the new method
The new quantisation method is named HIGGS (Hadamard Incoherence with Gaussian MSE-Optimal GridS). It enables the compression of neural networks without the need for additional data or computationally intensive parameter optimisation. This is especially useful in situations where there is not enough relevant data available to train the model. HIGGS strikes a balance between the quality, size, and complexity of the quantised models, making them suitable for use on a variety of devices.
The method has already been validated on the widely used Llama 3 and Qwen2.5 models. Experiments have shown that HIGGS outperforms all existing data-free quantisation methods, including NF4 (4-bit NormalFloat) and HQQ (Half-Quadratic Quantisation), in terms of both quality and model size.
Scientists from HSE University, the Massachusetts Institute of Technology (MIT), the Austrian Institute of Science and Technology (ISTA), and King Abdullah University of Science and Technology (KAUST, Saudi Arabia), all contributed to the development of the method.
The HIGGS method is already accessible to developers and researchers on Hugging Face and GitHub, with a research paper available on arXiv.
Response from the academic community, and other methods
The paper describing the new method has been accepted for presentation at one of the largest AI conferences in the world—the North American Chapter of the Association for Computational Linguistics (NAACL). The conference will be held from April 29 to May 4, 2025, in Albuquerque, New Mexico, USA, and Yandex will be among the attendees, along with other companies and universities such as Google, Microsoft Research, and Harvard University. The paper has been cited by Red Hat AI, an American software company, as well as Peking University, Hong Kong University of Science and Technology, Fudan University, and others.
Previously, scientists from Yandex presented 12 studies focused on LLM quantisation. The company aims to make the application of LLMs more efficient, less energy-consuming, and accessible to all developers and researchers. For example, the Yandex Research team has previously developed methods for compressing LLMs, which reduce computational costs by nearly eight times, while not significantly compromising the quality of the neural network’s responses. The team has also developed a solution that allows running a model with 8 billion parameters on a regular computer or smartphone through a browser interface, even without major computational power.
See also:
Scientists Develop Algorithm for More Reliable Processors in Data Centres
Researchers from HSE MIEM and Samara University have developed the LRF-3D algorithm to automatically bypass idle nodes in three-dimensional networks-on-chip. Thanks to its hierarchical architecture, the algorithm outperforms existing solutions in both speed and path accuracy, improving processor reliability for use in data centres, supercomputers, and AI computing. The source code and test results are publicly available.
Researchers Develop Method for Direct Generation of Regulatory DNA
Researchers at HSE University have developed a model for generating promoters and enhancers—DNA sequences that regulate gene activity. The model works directly with DNA nucleotides, without first transforming them into a continuous numerical representation. This solution could be useful for applications in synthetic biology and gene therapy. The study results were presented at the ICLR 2026 Workshop ‘Generative AI in Genomics (Gen^2): Barriers and Frontiers.’
Researchers at HSE University and Sber Train Neural Networks to Better Predict User Preferences
The HSE FCS AI and Digital Science Institute and Sber have introduced a new architecture for recommendation systems that combines two classes of models, enabling algorithms to better predict users’ interests and needs. A preprint of the paper has been published on arxiv.org and presented at Urban ML.
Physicists Discover What Happens Inside a Stable Vortex
Large vortices with characteristic spiral arms are often observed in the atmosphere and the ocean. Physicists from HSE University have explained how these structures form and why they retain their shape. The researchers found that velocities at points located along the same vortex arc remain correlated even over long distances. At the same time, this correlation weakens rapidly with increasing distance from the vortex centre. These differences help explain the formation of spiral arms and may improve models of atmospheric and oceanic currents. The findings have been published in Physical Review Fluids.
‘The Peak of Stupidity’ and ‘The Valley of Despair’: HSE Economists Propose an Explanation for the Dunning–Kruger Effect
The Dunning–Kruger effect, which describes a sharp surge in self-confidence among beginners followed by an equally rapid decline as they gain experience, can be explained by the nature of the learning process and the acquisition of new knowledge. This conclusion was reached by Andrey Vorchik of the HSE Faculty of Economic Sciences together with independent researcher Murat Mamyshev. They developed a mathematical model of learning and demonstrated how subjective confidence is formed and changes as knowledge accumulates, as well as how teachers can reduce the ‘valley of despair’ experienced by learners.
Toffee and Risk: Scientists Discover Why People Who Crave Sweets Make More Impulsive Choices
Having a sweet tooth may be linked not only to eating habits but also to the way people make decisions. Researchers at HSE University have found that people with a preference for sweet foods tend to behave more impulsively—not because they want immediate rewards, but because they are less willing to tolerate uncertainty. These findings may help improve treatments for addiction. The study findings have been published in Frontiers in Psychology.
Physicists Find a Way to Model Ion Parameters in Plasma in Seconds
Researchers from HSE University and the Moscow Institute of Physics and Technology (MIPT) have developed a set of simple analytical methods for calculating the properties of heavy ions in helium under the influence of a strong electric field. The new approach speeds up calculations of ion mobility and ion–molecule reaction rates by thousands of times while maintaining sufficient accuracy for plasma jet modelling. The findings have been published in the journal Physica Scripta.
Two Years of Growth or Decline: How to Choose an Investment Strategy
Economists from HSE University, together with colleagues from international universities, have analysed stock market movements over almost a century and proposed an investment strategy that could have delivered returns nearly twice as high as the market average. Their research suggests following a momentum strategy during periods of sustained market growth and switching to a value strategy after prolonged market declines. The study has been published in the Journal of Banking and Finance.
Researchers Reveal Link Between Attention and Communication Difficulties in Autism
Researchers at HSE University have examined how communication difficulties in children with autism are related to brain function. The findings show that not only language networks but also attention networks play an important role. The weaker the connections involved in maintaining focus and switching attention, the more pronounced communication difficulties were. The study has been published in European Child & Adolescent Psychiatry.
Scientists Discover Why Some People Wore Masks During COVID-19 While Others Did Not
Why do some people voluntarily follow new rules while others ignore them? Researchers at HSE University have found that the answer lies not so much in people's willingness to cooperate, as previously believed, but in their ability to empathise with others. Empathy proved to be the strongest predictor of whether people chose to wear face masks voluntarily during the COVID-19 pandemic. The findings have been published in Frontiers.