Researchers from Samsung Electronic Co. Ltd. have created a tiny artificial intelligence model that punches far above its weight on certain kinds of \u201creasoning\u201d tasks, challenging the industry\u2019s longheld logic that \u201cbigger means better\u201d.<\/p>\n
Released this week, the Tiny Recursive Model or TRM has just seven million parameters, far fewer than most other AI models, yet shows that it can outperform powerful large language models such as Google LLC\u2019s Gemini 2.5 Pro on tough reasoning puzzles such as Sudoku.<\/p>\n
Alexia Jolicoeur-Martineau, a senior researcher at the Samsung Advanced Institute of Technology AI Lab in Montreal<\/a>, published a paper on arXiv that demonstrates<\/a> how clever design can be more effective than simply increasing the number of parameters in AI models. It uses a special \u201crecursive reasoning\u201d process that enables it to think in \u201cloops\u201d, going over the same problem repeatedly in order to improve its answers.<\/p>\n The paper, titled \u201cLess is More: Recursive Reasoning with Tiny Networks,\u201d reveals how TRM was designed specifically to tackle logic puzzles and reasoning challenges. It\u2019s not able to chat with humans, write stories or create images like other models can. But its narrow focus means it can solve some really hard problems with greater accuracy than its much larger counterparts.<\/p>\n For instance, TRM achieved 87% accuracy on Sudoku-Extreme, a benchmark that challenges AI models to complete multiple Sudoku puzzles. It also racked up an 85% score on Maze-Hard, which tasks models with finding their way through complex mazes in the fastest time possible. And it scored 45% and 8% on the ARC-AGI-1 and ARC-AGI-2 benchmarks, which consist of more abstract reasoning puzzles designed to test for \u201cgeneral intelligence.\u201d<\/p>\n In each of these tasks, TRM outperformed much larger models. For instance, Gemini 2.5 Pro could only score 4.9% on the ARC-AGI-2 test, while OpenAI\u2019s o3-mini-high scored just 3%, DeepSeek Ltd.\u2019s R1 achieved just 1.3% and Anthropic PBC\u2019s Claude 3.7 could only muster a 0.7% score. TRM achieved this with less than 0.01% of the parameters used by the most powerful large language models.<\/p>\n New paper \ud83d\udcdc: Tiny Recursion Model (TRM) is a recursive reasoning approach with a tiny 7M parameters neural network that obtains 45% on ARC-AGI-1 and 8% on ARC-AGI-2, beating most LLMs.<\/p>\n Blog: https:\/\/t.co\/w5ZDsHDDPE<\/a><\/p>\n Code: https:\/\/t.co\/7UgKuD9Yll<\/a><\/p>\n Paper: https:\/\/t.co\/3m8ANhNMiw<\/a><\/p>\n \u2014 Alexia Jolicoeur-Martineau (@jm_alexia) October 7, 2025<\/a><\/p>\n Recursive reasoning loops<\/p>\n Rather than build a large neural network, Samsung\u2019s researchers looked at the possibility of using recursion, which is a technique humans can also use. Essentially, the model looks at its answer and asks itself, \u201cIs it any good? If not, can I come up with a better answer?\u201d. It then attempts to solve the puzzle again, refining its answer, and repeats this process until its satisfied.<\/p>\n To do this, TRM maintains two short-term memories \u2013 it remembers the current solution, and also creates a kind of scratchpad to jot down the intermediate steps it takes to try and improve on that. At each step, the model updates the scratchpad by reviewing the task, the current solutions and its previous notes, before generating an improved output based on that information.<\/p>\n It repeats this loop multiple times, gradually refining its answers, eliminating the need for lengthy reasoning chains that can only be handled by billions of parameters. Instead, only a small network of a few million parameters is required.<\/p>\n The researchers stated in the paper that TRM is programmed to \u201crecursively refine latent and output states without assuming convergence.\u201d What this means is that the model is not forced to settle on an answer too soon, but rather, allowed to keep repeating the loop until it\u2019s unable to improve its output any more. It uses an \u201cadaptive halting\u201d technique that allows it to figure out for itself when that happens, preventing it from running indefinitely. The model also employs deep supervision, which means it can obtain feedback at multiple steps of its reasoning process, instead of just at the end. This helps the model to learn more effectively, the authors said.<\/p>\n Less is more could be a big deal<\/p>\n Jolicoeur-Martineau said in a blog post<\/a> that the research is significant because it demonstrates that small, highly-targeted models can achieve excellent results on narrow, structured reasoning tasks, and it could be a significant development for the broader AI industry.<\/p>\n The obvious benefit is that it makes powerful AI systems more accessible. The biggest LLMs with billions or even trillions of parameters can only be run on enormous clusters of specialized and expensive graphics processing units. These consume vast amounts of energy, which means that only a handful of rich companies and well-funded universities can experiment with them. But a model like TRM, which only has a few million parameters, can be run on commodity hardware, with a much lower energy footprint.<\/p>\n < 500$, 4 H-100 for around 2days<\/p>\n \u2014 Alexia Jolicoeur-Martineau (@jm_alexia) October 7, 2025<\/a><\/p>\n It potentially opens the door for many more universities, startups and independent developers to experiment with advanced AI models and accelerate innovation.<\/p>\n That said, Jolicoeur-Martineau\u2019s team pointed out that their findings don\u2019t mean LLMs are obsolete. TRM can only operate effectively when handling well-defined grid problems, and is not suitable for open-ended, text-based or multimodal tasks. Nonetheless, it represents a promising development, and the researchers plan to conduct further experiments to try and adapt recursive learning models to new domains.<\/p>\n Image: SiliconANGLE\/Dreamina AI<\/p>\n Support our mission to keep content open and free by engaging with theCUBE community. Join theCUBE\u2019s Alumni Trust Network, where technology leaders connect, share intelligence and create opportunities.<\/p>\n 15M+ viewers of theCUBE videos, powering conversations across AI, cloud, cybersecurity and more About SiliconANGLE Media<\/p>\n SiliconANGLE Media is a recognized leader in digital media innovation, uniting breakthrough technology, strategic insights and real-time audience engagement. As the parent company of SiliconANGLE<\/a>, theCUBE Network<\/a>, theCUBE Research<\/a>, CUBE365<\/a>, theCUBE AI<\/a> and theCUBE SuperStudios \u2014 with flagship locations in Silicon Valley and the New York Stock Exchange \u2014 SiliconANGLE Media operates at the intersection of media, technology and AI.<\/p>\n Founded by tech visionaries John Furrier and Dave Vellante, SiliconANGLE Media has built a dynamic ecosystem of industry-leading digital media brands that reach 15+ million elite tech professionals. Our new proprietary theCUBE AI Video Cloud is breaking ground in audience interaction, leveraging theCUBEai.com neural network to help technology companies make data-driven decisions and stay at the forefront of industry conversations.<\/p>\n \t\t\t\t\t
\n11.4k+ theCUBE alumni \u2014 Connect with more than 11,400 tech and business leaders shaping the future through a unique trusted-based network.<\/p>\n