---
title: "Small Language Models are the Future of Agentic AI (2025)"
description: "They define SLMs loosely as models that run on consumer hardware fast enough to serve one user… 85 references, cited by 75 papers here."
authors: ["Peter Belcak", "Greg Heinrich", "Shizhe Diao", "Yonggan Fu", "Xin Dong", "Saurav Muralidharan", "Yingyan Celine Lin", "Pavlo Molchanov"]
year: 2025
venue: "arXiv:2506.02153"
source: "https://arxiv.org/abs/2506.02153"
doi: "10.48550/arxiv.2506.02153"
arxiv: "2506.02153"
licence: "CC BY 4.0 (arXiv)"
licence_url: "http://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/belcak-2025-small-language-models-future-agentic/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/belcak-2025-small-language-models-future-agentic/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2025](https://md.research.elamur.ai/free/en/papers/2025/) / Small Language Models are the Future of Agentic AI

# Small Language Models are the Future of Agentic AI

**Authors:** [Peter Belcak](https://md.research.elamur.ai/free/en/authors/peter-belcak/), Greg Heinrich, [Shizhe Diao](https://md.research.elamur.ai/free/en/authors/shizhe-diao/), [Yonggan Fu](https://md.research.elamur.ai/free/en/authors/yonggan-fu/), [Xin Dong](https://md.research.elamur.ai/free/en/authors/xin-dong/), [Saurav Muralidharan](https://md.research.elamur.ai/free/en/authors/saurav-muralidharan/), [Yingyan Celine Lin](https://md.research.elamur.ai/free/en/authors/yingyan-celine-lin/) and [Pavlo Molchanov](https://md.research.elamur.ai/free/en/authors/pavlo-molchanov/) · **Year:** 2025 · **Venue:** arXiv:2506.02153 · **Licence:** [CC BY 4.0 (arXiv)](http://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://arxiv.org/abs/2506.02153) · [PDF](https://arxiv.org/pdf/2506.02153) · [DOI 10.48550/arxiv.2506.02153](https://doi.org/10.48550/arxiv.2506.02153) · [arXiv 2506.02153](https://arxiv.org/abs/2506.02153) · [OpenAlex](https://openalex.org/W4415130727)

**Citations:** 17 citations (OpenAlex, 2026-10-14) · [75](https://md.research.elamur.ai/free/en/papers/belcak-2025-small-language-models-future-agentic/cited-by/) in our library, of the 149,439 papers whose bibliographies we read

## Summary
This position paper from NVIDIA Research and Georgia Tech challenges a common industry practice: sending every request an AI agent makes to one large, general-purpose LLM hosted in the cloud. The authors note that most subtasks inside deployed agents are narrow, repetitive and non-conversational. They argue that small language models (SLMs) are capable enough for this work, better suited operationally and cheaper. They define SLMs loosely as models that run on consumer hardware fast enough to serve one user, which in practice means most models under 10bn parameters.

The paper introduces no new data or model. Its case rests on published benchmark results for recent SLMs, the typical architecture of agentic systems, and the economics of inference. When open-ended dialogue or broad reasoning is required, the authors recommend heterogeneous agents: SLMs handle requests by default, and an LLM is called only occasionally. They also identify obstacles to adoption and outline a six-step procedure for migrating an agent from an LLM to fine-tuned SLM specialists. The work matters because even a partial shift could noticeably lower the cost and energy footprint of agent deployments.

## Key points
- **Core claims and evidence.** The authors make three claims: SLMs are sufficiently capable (V1), more operationally suitable (V2) and more economical (V3) for most agent calls. Supporting examples include:
  - Phi-2 (2.7bn) matches 30bn models on commonsense reasoning and code while running about 15× faster.
  - xLAM-2-8B surpasses GPT-4o and Claude 3.5 on tool calling.
  - DeepSeek-R1-Distill-Qwen-7B beats Claude-3.5-Sonnet-1022 and GPT-4o-0513.
  - Nemotron-H, SmolLM2, Hymba-1.5B and RETRO-7.5B are also cited.
- **Cost estimate.** By the authors' figures, serving a 7bn SLM costs 10–30× less in latency, energy and FLOPs than serving a 70–175bn LLM.
- **Adoption barriers.** The paper names three obstacles:
  - heavy existing investment in centralized LLM infrastructure (estimated at USD 57bn);
  - SLM development and evaluation that relies on generalist rather than agent-specific benchmarks;
  - low public awareness of SLMs.
- **LLM-to-SLM conversion algorithm.** The six steps are:
  1. secure logging of non-HCI agent calls;
  2. data curation and removal of PII/PHI, with roughly 10k–100k examples as a rule of thumb for fine-tuning;
  3. unsupervised clustering of tasks;
  4. SLM selection;
  5. fine-tuning with LoRA/QLoRA, full fine-tuning or knowledge distillation;
  6. periodic retraining.
- **Market context.** The agentic AI sector was valued at USD 5.2bn in late 2024 and is projected to reach nearly USD 200bn by 2034. The authors invite critique and promise to publish all responses at research.nvidia.com/labs/lpr/slm-agents.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

## Institutions

- [Georgia Institute of Technology](https://md.research.elamur.ai/free/en/universities/georgia-institute-of-technology/) — University, Atlanta, United States · 152 works · 27,763 citations
- [Nvidia](https://md.research.elamur.ai/free/en/companies/nvidia/) — Company, Santa Clara, United States · 37 works · 12,656 citations

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 18,614 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 17,965 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 17,450 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 17,296 works

🧭 Topics (1): [🧠 Language models & pretraining](https://md.research.elamur.ai/free/en/topics/language-models/) · 3,894 works

## Cite

Cite as: Peter Belcak, Greg Heinrich, Shizhe Diao, Yonggan Fu, Xin Dong, Saurav Muralidharan, Yingyan Celine Lin and Pavlo Molchanov. 2025. Small Language Models are the Future of Agentic AI. arXiv:2506.02153.

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@misc{belcak2025small,
  title = {{Small Language Models are the Future of Agentic AI}},
  author = {Peter Belcak, Greg Heinrich, Shizhe Diao, Yonggan Fu, Xin Dong, Saurav Muralidharan, Yingyan Celine Lin and Pavlo Molchanov},
  year = {2025},
  doi = {10.48550/arxiv.2506.02153},
  eprint = {2506.02153},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2506.02153}
}
```

- [← previous Qwen-Image Technical Report](https://md.research.elamur.ai/free/en/papers/wu-2025-qwen-image-technical-report/)
- [next → Vision-R1: Incentivizing Reasoning Capability in Multimodal Large Language Models](https://md.research.elamur.ai/free/en/papers/huang-2025-vision-r1-incentivizing-reasoning-capability/)
- [↑ back to papers of 2025](https://md.research.elamur.ai/free/en/papers/2025/)

## References

85 references, 23 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Aashima. Small language models vs. llms: Finding the right fit for your needs, October 2024. Accessed: 2025-05-09. [find](https://scholar.google.com/scholar?q=Aashima%2E+Small+language+models+vs%2E+llms%3A+Finding+the+right+fit+for+your+needs%2C+October+2024%2E+Accessed%3A+2025%2D05%2D09%2E)
2. ABBYY. Small language models vs. large language models, November 2024. Accessed: 2025-05-09. [find](https://scholar.google.com/scholar?q=ABBYY%2E+Small+language+models+vs%2E+large+language+models%2C+November+2024%2E+Accessed%3A+2025%2D05%2D09%2E)
3. Marah Abdin, Jyoti Aneja, Hany Awadalla, Ahmed Awadallah, Ammar Ahmad Awan, Nguyen Bach, Amit Bahree, Arash Bakhtiari, Jianmin Bao, Harkirat Behl, et al. [Phi-3 technical report: A highly capable language model locally on your phone](https://md.research.elamur.ai/free/en/papers/abdin-2024-phi-3-technical-report-highly/). arXiv preprint arXiv:2404.14219, 2024.
4. Adyog. The economics of ai training and inference: How deepseek broke the cost curve, February 2025. Accessed: 2025-05-09. [find](https://scholar.google.com/scholar?q=Adyog%2E+The+economics+of+ai+training+and+inference%3A+How+deepseek+broke+the+cost+curve%2C+February+2025%2E+Accessed%3A+2025%2D05%2D09%2E)
5. Ishika Agarwal, Krishnateja Killamsetty, Lucian Popa, and Marina Danilevksy. Delift: Data efficient language model instruction fine tuning. arXiv preprint arXiv:2411.04425, 2024. — [arXiv:2411.04425](https://arxiv.org/abs/2411.04425) [find](https://scholar.google.com/scholar?q=Ishika+Agarwal%2C+Krishnateja+Killamsetty%2C+Lucian+Popa%2C+and+Marina+Danilevksy%2E+Delift%3A+Data+efficient+language+model+instruction+fine+tuning%2E+arXiv+preprint+arXiv%3A2411%2E04425%2C+2024%2E+%E2%80%94+%5BarXiv%3A2411%2E04425%5D%28)
6. Loubna Ben Allal, Anton Lozhkov, Elie Bakouch, Gabriel Martín Blázquez, Guilherme Penedo, Lewis Tunstall, Andrés Marafioti, Hynek Kydlíček, Agustín Piqueres Lajarín, Vaibhav Srivastav, Joshua Lochner, Caleb Fahlgren, Xuan-Son Nguyen, Clémentine Fourrier, Ben Burtenshaw, Hugo Larcher, Haojun Zhao, C… — [SmolLM2: When Smol Goes Big -- Data-Centric Training of a Small Language Model](https://md.research.elamur.ai/free/en/papers/allal-2025-smollm2-smol-goes-big-data/)
7. Peter Belcak, Greg Heinrich, Jan Kautz, and Pavlo Molchanov. Minifinetuning: Low-data generation domain adaptation through corrective self-distillation. arXiv preprint arXiv:2506.15702, 2025. — [arXiv:2506.15702](https://arxiv.org/abs/2506.15702) [find](https://scholar.google.com/scholar?q=Peter+Belcak%2C+Greg+Heinrich%2C+Jan+Kautz%2C+and+Pavlo+Molchanov%2E+Minifinetuning%3A+Low%2Ddata+generation+domain+adaptation+through+corrective+self%2Ddistillation%2E+arXiv+preprint+arXiv%3A2506%2E15702%2C+2025%2E+%E2%80%94+%5BarXiv)
8. Peter Belcak and Roger Wattenhofer. Tiny transformers excel at sentence compression. arXiv preprint arXiv:2410.23510, 2024. — [arXiv:2410.23510](https://arxiv.org/abs/2410.23510) [find](https://scholar.google.com/scholar?q=Peter+Belcak+and+Roger+Wattenhofer%2E+Tiny+transformers+excel+at+sentence+compression%2E+arXiv+preprint+arXiv%3A2410%2E23510%2C+2024%2E+%E2%80%94+%5BarXiv%3A2410%2E23510%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F2410%2E23510%29)
9. Aaron Blakeman, Aarti Basant, Abhinav Khattar, Adithya Renduchintala, Akhiad Bercovich, Aleksander Ficek, Alexis Bjorlin, Ali Taghibakhshi, Amala Sanjay Deshmukh, Ameya Sunil Mahabaleshwarkar, et al. [Nemotron-h: A family of accurate and efficient hybrid mamba-transformer models](https://md.research.elamur.ai/free/en/papers/blakeman-2025-nemotron-h-family-accurate-efficient/). arXiv preprint arXiv…
10. Sebastian Borgeaud, Arthur Mensch, Jordan Hoffmann, Trevor Cai, Eliza Rutherford, Katie Millican, George van den Driessche, Bogdan Damoc, Aidan Clark, Jan Kramár, et al. Improving language models by retrieving from trillions of tokens. arXiv preprint arXiv:2112.04426, 2022. [find](https://scholar.google.com/scholar?q=Sebastian+Borgeaud%2C+Arthur+Mensch%2C+Jordan+Hoffmann%2C+Trevor+Cai%2C+Eliza+Rutherford%2C+Katie+Millican%2C+George+van+den+Driessche%2C+Bogdan+Damoc%2C+Aidan+Clark%2C+Jan+Kram%C3%A1r%2C+et+al%2E+Improving+language+models+by+r)
11. Michael Brennan, Sadia Afroz, and Rachel Greenstadt. Adversarial stylometry: Circumventing authorship recognition to preserve privacy and anonymity. ACM Transactions on Information and System Security (TISSEC), 15(3):1–22, 2012. [find](https://scholar.google.com/scholar?q=Michael+Brennan%2C+Sadia+Afroz%2C+and+Rachel+Greenstadt%2E+Adversarial+stylometry%3A+Circumventing+authorship+recognition+to+preserve+privacy+and+anonymity%2E+ACM+Transactions+on+Information+and+System+Security)
12. Ruisi Cai, Saurav Muralidharan, Greg Heinrich, Hongxu Yin, Zhangyang Wang, Jan Kautz, and Pavlo Molchanov. Flextron: Many-in-one flexible large language model. In Proceedings of the 41st International Conference on Machine Learning (ICML 2024), 2024. [find](https://scholar.google.com/scholar?q=Ruisi+Cai%2C+Saurav+Muralidharan%2C+Greg+Heinrich%2C+Hongxu+Yin%2C+Zhangyang+Wang%2C+Jan+Kautz%2C+and+Pavlo+Molchanov%2E+Flextron%3A+Many%2Din%2Done+flexible+large+language+model%2E+In+Proceedings+of+the+41st+International)
13. Michael Chui, Bryce Hall, Helen Mayhew, Alex Singla, and Alexander Sukharevsky. The state of ai in 2022—and a half decade in review, December 2022. Accessed: 2025-05-09. [find](https://scholar.google.com/scholar?q=Michael+Chui%2C+Bryce+Hall%2C+Helen+Mayhew%2C+Alex+Singla%2C+and+Alexander+Sukharevsky%2E+The+state+of+ai+in+2022%E2%80%94and+a+half+decade+in+review%2C+December+2022%2E+Accessed%3A+2025%2D05%2D09%2E)
14. Cloudera, Inc. 96% of enterprises are expanding use of ai agents, according to latest data from cloudera, April 2025. Accessed: 2025-05-08. [find](https://scholar.google.com/scholar?q=Cloudera%2C+Inc%2E+96%25+of+enterprises+are+expanding+use+of+ai+agents%2C+according+to+latest+data+from+cloudera%2C+April+2025%2E+Accessed%3A+2025%2D05%2D08%2E)
15. Planck Collaboration et al. Planck 2018 results. vi. cosmological parameters. Astronomy & Astrophysics, 641:A6, 2020. [find](https://scholar.google.com/scholar?q=Planck+Collaboration+et+al%2E+Planck+2018+results%2E+vi%2E+cosmological+parameters%2E+Astronomy+%26+Astrophysics%2C+641%3AA6%2C+2020%2E)
16. Colliers. 2025 data center marketplace: Balancing unprecedented opportunity with strategic risk. U.s. research report, Colliers, 2025. [find](https://scholar.google.com/scholar?q=Colliers%2E+2025+data+center+marketplace%3A+Balancing+unprecedented+opportunity+with+strategic+risk%2E+U%2Es%2E+research+report%2C+Colliers%2C+2025%2E)
17. DAIR.AI. Llm agents, April 2024. Accessed: 2025-05-08. [find](https://scholar.google.com/scholar?q=DAIR%2EAI%2E+Llm+agents%2C+April+2024%2E+Accessed%3A+2025%2D05%2D08%2E)
18. Badhan Chandra Das, M Hadi Amini, and Yanzhao Wu. [Security and privacy challenges of large language models: A survey](https://md.research.elamur.ai/free/en/papers/das-2024-security-privacy-challenges-large-language/). ACM Computing Surveys, 57(6):1–39, 2025.
19. DeepSeek-AI. [Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning](https://md.research.elamur.ai/free/en/papers/deepseek-ai-2025-deepseek-r1-incentivizing-reasoning-capability/), 2025.
20. Tim Dettmers, Artidoro Pagnoni, Ari Holtzman, and Luke Zettlemoyer. [Qlora: Efficient finetuning of quantized llms](https://md.research.elamur.ai/free/en/papers/dettmers-2023-qlora-efficient-finetuning-quantized-llms/). Advances in neural information processing systems, 36:10088– 10115, 2023.
21. Shizhe Diao, Yu Yang, Yonggan Fu, Xin Dong, Dan Su, Markus Kliegl, Zijia Chen, Peter Belcak, Yoshi Suhara, Hongxu Yin, et al. Climb: Clustering-based iterative data mixture bootstrapping for language model pre-training. arXiv preprint arXiv:2504.13161, 2025. — [arXiv:2504.13161](https://arxiv.org/abs/2504.13161) [find](https://scholar.google.com/scholar?q=Shizhe+Diao%2C+Yu+Yang%2C+Yonggan+Fu%2C+Xin+Dong%2C+Dan+Su%2C+Markus+Kliegl%2C+Zijia+Chen%2C+Peter+Belcak%2C+Yoshi+Suhara%2C+Hongxu+Yin%2C+et+al%2E+Climb%3A+Clustering%2Dbased+iterative+data+mixture+bootstrapping+for+language+)
22. Ning Ding, Yujia Qin, Guang Yang, Fuchao Wei, Zonghan Yang, Yusheng Su, Shengding Hu, Yulin Chen, Chi-Min Chan, Weize Chen, et al. [Parameter-efficient fine-tuning of large-scale pre-trained language models](https://md.research.elamur.ai/free/en/papers/ding-2023-parameter-efficient-fine-tuning-large/). Nature Machine Intelligence, 5(3):220–235, 2023.
23. Xin Dong, Yonggan Fu, Shizhe Diao, Wonmin Byeon, Zijia Chen, Ameya Sunil Mahabaleshwarkar, Shih-Yang Liu, Matthijs Van Keirsbilck, Min-Hung Chen, Yoshi Suhara, et al. [Hymba: A hybrid-head architecture for small language models](https://md.research.elamur.ai/free/en/papers/dong-2024-hymba-hybrid-head-architecture-small/). arXiv preprint arXiv:2411.13676, 2024.
24. Amr Elmeleegy et al. Introducing nvidia dynamo, a low-latency distributed inference framework for scaling reasoning ai models, March 2025. NVIDIA Technical Blog. [find](https://scholar.google.com/scholar?q=Amr+Elmeleegy+et+al%2E+Introducing+nvidia+dynamo%2C+a+low%2Dlatency+distributed+inference+framework+for+scaling+reasoning+ai+models%2C+March+2025%2E+NVIDIA+Technical+Blog%2E)
25. Henry Evans. Llms vs. slms: Balancing comprehensiveness and smart resource-saving, April 2025. Accessed: 2025-05-09. [find](https://scholar.google.com/scholar?q=Henry+Evans%2E+Llms+vs%2E+slms%3A+Balancing+comprehensiveness+and+smart+resource%2Dsaving%2C+April+2025%2E+Accessed%3A+2025%2D05%2D09%2E)
26. Barbara A Ferguson, Timothy A Dreisbach, Catherine G Parks, Gregory M Filip, and Craig L Schmitt. Coarse-scale population structure of pathogenic armillaria species in a mixed-conifer forest in the blue mountains of northeast oregon. Canadian Journal of Forest Research, 33(4):612–623, 2003. [find](https://scholar.google.com/scholar?q=Barbara+A+Ferguson%2C+Timothy+A+Dreisbach%2C+Catherine+G+Parks%2C+Gregory+M+Filip%2C+and+Craig+L+Schmitt%2E+Coarse%2Dscale+population+structure+of+pathogenic+armillaria+species+in+a+mixed%2Dconifer+forest+in+the+bl)
27. Yonggan Fu, Zhongzhi Yu, Junwei Li, Jiayi Qian, Yongan Zhang, Xiangchi Yuan, Dachuan Shi, Roman Yakunin, and Yingyan Celine Lin. Amoeballm: Constructing any-shape large language models for efficient and instant deployment. In Proceedings of the 38th Annual Conference on Neural Information Processing… [find](https://scholar.google.com/scholar?q=Yonggan+Fu%2C+Zhongzhi+Yu%2C+Junwei+Li%2C+Jiayi+Qian%2C+Yongan+Zhang%2C+Xiangchi+Yuan%2C+Dachuan+Shi%2C+Roman+Yakunin%2C+and+Yingyan+Celine+Lin%2E+Amoeballm%3A+Constructing+any%2Dshape+large+language+models+for+efficient+a)
28. google. GitHub - google/A2A: An open protocol enabling communication and interoperability between opaque agentic applications. [find](https://scholar.google.com/scholar?q=google%2E+GitHub+%2D+google%2FA2A%3A+An+open+protocol+enabling+communication+and+interoperability+between+opaque+agentic+applications%2E)
29. David Gu, Peter Belcak, and Roger Wattenhofer. Text compression for efficient language generation. arXiv preprint arXiv:2503.11426, 2025. [find](https://scholar.google.com/scholar?q=David+Gu%2C+Peter+Belcak%2C+and+Roger+Wattenhofer%2E+Text+compression+for+efficient+language+generation%2E+arXiv+preprint+arXiv%3A2503%2E11426%2C+2025%2E)
30. Harrison Clarke. Large language models vs. small language models, March 2024. Accessed: 2025-05-09. [find](https://scholar.google.com/scholar?q=Harrison+Clarke%2E+Large+language+models+vs%2E+small+language+models%2C+March+2024%2E+Accessed%3A+2025%2D05%2D09%2E)
31. Danny Hernandez, Jared Kaplan, Tom Henighan, and Sam McCandlish. [Scaling laws for transfer](https://md.research.elamur.ai/free/en/papers/hernandez-2021-scaling-laws-transfer/). arXiv preprint arXiv:2102.01293, 2021.
32. Jordan Hoffmann, Sebastian Borgeaud, Arthur Mensch, Elena Buchatskaya, Trevor Cai, Eliza Rutherford, Diego de Las Casas, Lisa Anne Hendricks, Johannes Welbl, Aidan Clark, et al. [Training compute-optimal large language models](https://md.research.elamur.ai/free/en/papers/hoffmann-2022-training-compute-optimal-large-language/). arXiv preprint arXiv:2203.15556, 2022.
33. Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. [Lora: Low-rank adaptation of large language models](https://md.research.elamur.ai/free/en/papers/hu-2021-lora-low-rank-adaptation-large/). arxiv 2021. arXiv preprint arXiv:2106.09685, 2021.
34. Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, Weizhu Chen, et al. [Lora: Low-rank adaptation of large language models](https://md.research.elamur.ai/free/en/papers/hu-2021-lora-low-rank-adaptation-large/). ICLR, 1(2):3, 2022.
35. Shaohan Huang, Furu Wei, Lei Cui, Xingxing Zhang, and Ming Zhou. Unsupervised fine-tuning for text clustering. In Proceedings of the 28th international conference on computational linguistics, pages 5530–5534, 2020. [find](https://scholar.google.com/scholar?q=Shaohan+Huang%2C+Furu+Wei%2C+Lei+Cui%2C+Xingxing+Zhang%2C+and+Ming+Zhou%2E+Unsupervised+fine%2Dtuning+for+text+clustering%2E+In+Proceedings+of+the+28th+international+conference+on+computational+linguistics%2C+pages+5)
36. Invisible Technologies. How small language models can outperform llms, March 2025. Accessed: 2025-05-21. [find](https://scholar.google.com/scholar?q=Invisible+Technologies%2E+How+small+language+models+can+outperform+llms%2C+March+2025%2E+Accessed%3A+2025%2D05%2D21%2E)
37. Mojan Javaheripi and Sébastien Bubeck. Phi-2: The surprising power of small language models, 2023. Microsoft Research Blog. [find](https://scholar.google.com/scholar?q=Mojan+Javaheripi+and+S%C3%A9bastien+Bubeck%2E+Phi%2D2%3A+The+surprising+power+of+small+language+models%2C+2023%2E+Microsoft+Research+Blog%2E)
38. Andreas Jungherr. Artificial intelligence and democracy: A conceptual framework. Social media+ society, 9(3):20563051231186353, 2023. [find](https://scholar.google.com/scholar?q=Andreas+Jungherr%2E+Artificial+intelligence+and+democracy%3A+A+conceptual+framework%2E+Social+media%2B+society%2C+9%283%29%3A20563051231186353%2C+2023%2E)
39. Aviv Kaufmann. Understanding the total cost of inferencing large language models. Technical report, Enterprise Strategy Group, April 2024. Commissioned by Dell Technologies. Accessed: 2025-05-09. [find](https://scholar.google.com/scholar?q=Aviv+Kaufmann%2E+Understanding+the+total+cost+of+inferencing+large+language+models%2E+Technical+report%2C+Enterprise+Strategy+Group%2C+April+2024%2E+Commissioned+by+Dell+Technologies%2E+Accessed%3A+2025%2D05%2D09%2E)
40. Sneha Kudugunta, Aditya Kusupati, Tim Dettmers, Kaifeng Chen, Inderjit Dhillon, Yulia Tsvetkov, Hannaneh Hajishirzi, Sham Kakade, Ali Farhadi, Prateek Jain, et al. Matformer: Nested transformer for elastic inference. arXiv preprint arXiv:2310.07707, 2023. — [arXiv:2310.07707](https://arxiv.org/abs/2310.07707) [find](https://scholar.google.com/scholar?q=Sneha+Kudugunta%2C+Aditya+Kusupati%2C+Tim+Dettmers%2C+Kaifeng+Chen%2C+Inderjit+Dhillon%2C+Yulia+Tsvetkov%2C+Hannaneh+Hajishirzi%2C+Sham+Kakade%2C+Ali+Farhadi%2C+Prateek+Jain%2C+et+al%2E+Matformer%3A+Nested+transformer+for+el)
41. Akshi Kumar. From large to small: The rise of small language models (slms) in text analytics. 2025. [find](https://scholar.google.com/scholar?q=Akshi+Kumar%2E+From+large+to+small%3A+The+rise+of+small+language+models+%28slms%29+in+text+analytics%2E+2025%2E)
42. Luying Liu, Jianchu Kang, Jing Yu, and Zhongliang Wang. A comparative study on unsupervised feature selection methods for text clustering. In 2005 International Conference on Natural Language Processing and Knowledge Engineering, pages 597–601. IEEE, 2005. [find](https://scholar.google.com/scholar?q=Luying+Liu%2C+Jianchu+Kang%2C+Jing+Yu%2C+and+Zhongliang+Wang%2E+A+comparative+study+on+unsupervised+feature+selection+methods+for+text+clustering%2E+In+2005+International+Conference+on+Natural+Language+Processi)
43. Shih-Yang Liu, Chien-Yi Wang, Hongxu Yin, Pavlo Molchanov, Yu-Chiang Frank Wang, Kwang-Ting Cheng, and Min-Hung Chen. Dora: Weight-decomposed low-rank adaptation. arXiv preprint arXiv:2402.09353, 2024. [find](https://scholar.google.com/scholar?q=Shih%2DYang+Liu%2C+Chien%2DYi+Wang%2C+Hongxu+Yin%2C+Pavlo+Molchanov%2C+Yu%2DChiang+Frank+Wang%2C+Kwang%2DTing+Cheng%2C+and+Min%2DHung+Chen%2E+Dora%3A+Weight%2Ddecomposed+low%2Drank+adaptation%2E+arXiv+preprint+arXiv%3A2402%2E09353%2C+2024)
44. Zichang Liu, Jue Wang, Tri Dao, Tianyi Zhou, Binhang Yuan, Zhao Song, Anshumali Shrivastava, Ce Zhang, Yuandong Tian, Christopher Re, et al. [Deja vu: Contextual sparsity for efficient llms at inference time](https://md.research.elamur.ai/free/en/papers/liu-2023-deja-vu-contextual-sparsity-efficient/). In International Conference on Machine Learning, pages 22137–22176. PMLR, 2023.
45. Nunzio Lore, Sepehr Ilami, and Babak Heydari. Large model strategic thinking, small model efficiency: transferring theory of mind in large language models. arXiv preprint arXiv:2408.05241, 2024. — [arXiv:2408.05241](https://arxiv.org/abs/2408.05241) [find](https://scholar.google.com/scholar?q=Nunzio+Lore%2C+Sepehr+Ilami%2C+and+Babak+Heydari%2E+Large+model+strategic+thinking%2C+small+model+efficiency%3A+transferring+theory+of+mind+in+large+language+models%2E+arXiv+preprint+arXiv%3A2408%2E05241%2C+2024%2E+%E2%80%94+%5Bar)
46. Jeff Loucks, Gillian Crossan, Baris Sarer, China Widener, and Ariane Bucaille. Autonomous generative ai agents: Under development. Deloitte Insights, November 2024. Accessed: 2025-05-08. [find](https://scholar.google.com/scholar?q=Jeff+Loucks%2C+Gillian+Crossan%2C+Baris+Sarer%2C+China+Widener%2C+and+Ariane+Bucaille%2E+Autonomous+generative+ai+agents%3A+Under+development%2E+Deloitte+Insights%2C+November+2024%2E+Accessed%3A+2025%2D05%2D08%2E)
47. Zhenyan Lu, Xiang Li, Dongqi Cai, Rongjie Yi, Fangming Liu, Xiwen Zhang, Nicholas D Lane, and Mengwei Xu. [Small language models: Survey, measurements, and insights](https://md.research.elamur.ai/free/en/papers/lu-2024-small-language-models-survey-measurements/). arXiv preprint arXiv:2409.15790, 2024.
48. Junyu Luo, Weizhi Zhang, Ye Yuan, Yusheng Zhao, Junwei Yang, Yiyang Gu, Bohan Wu, Binqi Chen, Ziyue Qiao, Qingqing Long, et al. [Large language model agent: A survey on methodology, applications and challenges](https://md.research.elamur.ai/free/en/papers/luo-2025-large-language-model-agent-survey/). arXiv preprint arXiv:2503.21460, 2025.
49. Georgina M Mace, Paul H Harvey, and Timothy H Clutton-Brock. Brain size and ecology in small mammals. Journal of Zoology, 193(3):333–354, 1981. [find](https://scholar.google.com/scholar?q=Georgina+M+Mace%2C+Paul+H+Harvey%2C+and+Timothy+H+Clutton%2DBrock%2E+Brain+size+and+ecology+in+small+mammals%2E+Journal+of+Zoology%2C+193%283%29%3A333%E2%80%93354%2C+1981%2E)
50. Tobias Mann. A closer look at dynamo, nvidia’s ’operating system’ for ai inference, March 2025. Accessed: 2025-05-09. [find](https://scholar.google.com/scholar?q=Tobias+Mann%2E+A+closer+look+at+dynamo%2C+nvidia%E2%80%99s+%E2%80%99operating+system%E2%80%99+for+ai+inference%2C+March+2025%2E+Accessed%3A+2025%2D05%2D09%2E)
51. Market.us. Global agentic ai market size, share analysis by product type, agent role, agent system, end user, region and companies – industry segment outlook, market assessment, competition scenario, trends and forecast 2025–2034, March 2025. Accessed: 2025-05-08. [find](https://scholar.google.com/scholar?q=Market%2Eus%2E+Global+agentic+ai+market+size%2C+share+analysis+by+product+type%2C+agent+role%2C+agent+system%2C+end+user%2C+region+and+companies+%E2%80%93+industry+segment+outlook%2C+market+assessment%2C+competition+scenario%2C+)
52. Tula Masterman, Sandi Besen, Mason Sawtell, and Alex Chao. The landscape of emerging ai agent architectures for reasoning, planning, and tool calling: A survey. arXiv preprint arXiv:2404.11584, 2024. [find](https://scholar.google.com/scholar?q=Tula+Masterman%2C+Sandi+Besen%2C+Mason+Sawtell%2C+and+Alex+Chao%2E+The+landscape+of+emerging+ai+agent+architectures+for+reasoning%2C+planning%2C+and+tool+calling%3A+A+survey%2E+arXiv+preprint+arXiv%3A2404%2E11584%2C+2024%2E)
53. Sourabh Mehta. How much energy do llms consume? unveiling the power behind ai, July 2024. Accessed: 2025-05-21. [find](https://scholar.google.com/scholar?q=Sourabh+Mehta%2E+How+much+energy+do+llms+consume%3F+unveiling+the+power+behind+ai%2C+July+2024%2E+Accessed%3A+2025%2D05%2D21%2E)
54. Meta Platforms, Inc. Model cards and prompt formats: Llama 3.3, April 2025. Accessed: 2025-05-08. [find](https://scholar.google.com/scholar?q=Meta+Platforms%2C+Inc%2E+Model+cards+and+prompt+formats%3A+Llama+3%2E3%2C+April+2025%2E+Accessed%3A+2025%2D05%2D08%2E)
55. Metomic. Understanding ai agents & data security, 2025. Accessed: 2025-05-13. [find](https://scholar.google.com/scholar?q=Metomic%2E+Understanding+ai+agents+%26+data+security%2C+2025%2E+Accessed%3A+2025%2D05%2D13%2E)
56. Erik Miehling, Karthikeyan Natesan Ramamurthy, Kush R Varshney, Matthew Riemer, Djallel Bouneffouf, John T Richards, Amit Dhurandhar, Elizabeth M Daly, Michael Hind, Prasanna Sattigeri, et al. Agentic ai needs a systems theory. arXiv preprint arXiv:2503.00237, 2025. — [arXiv:2503.00237](https://arxiv.org/abs/2503.00237) [find](https://scholar.google.com/scholar?q=Erik+Miehling%2C+Karthikeyan+Natesan+Ramamurthy%2C+Kush+R+Varshney%2C+Matthew+Riemer%2C+Djallel+Bouneffouf%2C+John+T+Richards%2C+Amit+Dhurandhar%2C+Elizabeth+M+Daly%2C+Michael+Hind%2C+Prasanna+Sattigeri%2C+et+al%2E+Agentic)
57. Morgan Stanley. Genai revenue growth and profitability, April 2025. Accessed: 2025-05-08. [find](https://scholar.google.com/scholar?q=Morgan+Stanley%2E+Genai+revenue+growth+and+profitability%2C+April+2025%2E+Accessed%3A+2025%2D05%2D08%2E)
58. Humza Naveed, Asad Ullah Khan, Shi Qiu, Muhammad Saqib, Saeed Anwar, Muhammad Usman, Naveed Akhtar, Nick Barnes, and Ajmal Mian. [A comprehensive overview of large language models](https://md.research.elamur.ai/free/en/papers/naveed-2023-comprehensive-overview-large-language-models/). arXiv preprint arXiv:2307.06435, 2023.
59. NVIDIA. Chatrtx, 2024. NVIDIA AI Product. [find](https://scholar.google.com/scholar?q=NVIDIA%2E+Chatrtx%2C+2024%2E+NVIDIA+AI+Product%2E)
60. NVIDIA. Nvidia dynamo: A datacenter scale distributed inference serving framework. [https://github.com/ai-dynamo/dynamo](https://github.com/ai-dynamo/dynamo), 2025. Accessed: 2025-05-09. [↗](https://github.com/ai-dynamo/dynamo)
61. Felipe Maia Polo, Lucas Weber, Leshem Choshen, Yuekai Sun, Gongjun Xu, and Mikhail Yurochkin. tinybenchmarks: evaluating llms with fewer examples. arXiv preprint arXiv:2402.14992, 2024. [find](https://scholar.google.com/scholar?q=Felipe+Maia+Polo%2C+Lucas+Weber%2C+Leshem+Choshen%2C+Yuekai+Sun%2C+Gongjun+Xu%2C+and+Mikhail+Yurochkin%2E+tinybenchmarks%3A+evaluating+llms+with+fewer+examples%2E+arXiv+preprint+arXiv%3A2402%2E14992%2C+2024%2E)
62. Lakshmi Radhakrishnan, Gundolf Schenk, Kathleen Muenzen, Boris Oskotsky, Habibeh Ashouri Choshali, Thomas Plunkett, Sharat Israni, and Atul J Butte. A certified de-identification system for all clinical text documents for information extraction at scale. JAMIA open, 6(3):ooad045, 2023. [find](https://scholar.google.com/scholar?q=Lakshmi+Radhakrishnan%2C+Gundolf+Schenk%2C+Kathleen+Muenzen%2C+Boris+Oskotsky%2C+Habibeh+Ashouri+Choshali%2C+Thomas+Plunkett%2C+Sharat+Israni%2C+and+Atul+J+Butte%2E+A+certified+de%2Didentification+system+for+all+clinic)
63. Martin J Rees. Before the Beginning: Our Universe and Others. Addison-Wesley, 1997. [find](https://scholar.google.com/scholar?q=Martin+J+Rees%2E+Before+the+Beginning%3A+Our+Universe+and+Others%2E+Addison%2DWesley%2C+1997%2E)
64. Judith Sáinz-Pardo Díaz and Álvaro López García. An open source python library for anonymizing sensitive data. Scientific data, 11(1):1289, 2024. [find](https://scholar.google.com/scholar?q=Judith+S%C3%A1inz%2DPardo+D%C3%ADaz+and+%C3%81lvaro+L%C3%B3pez+Garc%C3%ADa%2E+An+open+source+python+library+for+anonymizing+sensitive+data%2E+Scientific+data%2C+11%281%29%3A1289%2C+2024%2E)
65. Timo Schick, Jane Dwivedi-Yu, Roberto Dessì, Roberta Raileanu, Maria Lomeli, Luke Zettlemoyer, Nicola Cancedda, and Thomas Scialom. [Toolformer: Language models can teach themselves to use tools](https://md.research.elamur.ai/free/en/papers/schick-2023-toolformer-language-models-can-teach/). In Advances in Neural Information Processing Systems (NeurIPS), 2023.
66. J William Schopf. Microfossils of the early archean apex chert: New evidence of the antiquity of life. Science, 260(5108):640–646, 1993. [find](https://scholar.google.com/scholar?q=J+William+Schopf%2E+Microfossils+of+the+early+archean+apex+chert%3A+New+evidence+of+the+antiquity+of+life%2E+Science%2C+260%285108%29%3A640%E2%80%93646%2C+1993%2E)
67. Tanya Seda. Cloud llm cost model: Breakdown for mid-market businesses, 2024. Accessed: 2025-05-09. [find](https://scholar.google.com/scholar?q=Tanya+Seda%2E+Cloud+llm+cost+model%3A+Breakdown+for+mid%2Dmarket+businesses%2C+2024%2E+Accessed%3A+2025%2D05%2D09%2E)
68. Olivia Shone. Explore ai models: Key differences between small language models and large language models, November 2024. Accessed: 2025-05-21. [find](https://scholar.google.com/scholar?q=Olivia+Shone%2E+Explore+ai+models%3A+Key+differences+between+small+language+models+and+large+language+models%2C+November+2024%2E+Accessed%3A+2025%2D05%2D21%2E)
69. Yixin Song, Zeyu Mi, Haotong Xie, and Haibo Chen. [Powerinfer: Fast large language model serving with a consumer-grade gpu](https://md.research.elamur.ai/free/en/papers/song-2023-powerinfer-fast-large-language-model/). In Proceedings of the ACM SIGOPS 30th Symposium on Operating Systems Principles, pages 590–606, 2024.
70. Shreyas Subramanian, Vikram Elango, and Mecit Gungor. Small language models (slms) can still pack a punch: A survey. arXiv preprint arXiv:2501.05465, 2025. — [arXiv:2501.05465](https://arxiv.org/abs/2501.05465) [find](https://scholar.google.com/scholar?q=Shreyas+Subramanian%2C+Vikram+Elango%2C+and+Mecit+Gungor%2E+Small+language+models+%28slms%29+can+still+pack+a+punch%3A+A+survey%2E+arXiv+preprint+arXiv%3A2501%2E05465%2C+2025%2E+%E2%80%94+%5BarXiv%3A2501%2E05465%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F2)
71. Synergy Technical. Small language models vs. large language models, 2025. Accessed: 2025-05-09. [find](https://scholar.google.com/scholar?q=Synergy+Technical%2E+Small+language+models+vs%2E+large+language+models%2C+2025%2E+Accessed%3A+2025%2D05%2D09%2E)
72. Brian G. Thamm. Trustworthy and secure ai: How small language models strengthen data security. Service Contractor Magazine, October 2024. Accessed: 2025-05-08. [find](https://scholar.google.com/scholar?q=Brian+G%2E+Thamm%2E+Trustworthy+and+secure+ai%3A+How+small+language+models+strengthen+data+security%2E+Service+Contractor+Magazine%2C+October+2024%2E+Accessed%3A+2025%2D05%2D08%2E)
73. Fali Wang, Zhiwei Zhang, Xianren Zhang, Zongyu Wu, Tzuhao Mo, Qiuhao Lu, Wanjing Wang, Rui Li, Junjie Xu, Xianfeng Tang, et al. [A comprehensive survey of small language models in the era of large language models: Techniques, enhancements, applications, collaboration with llms, and trustworthiness](https://md.research.elamur.ai/free/en/papers/wang-2024-comprehensive-survey-small-language-models/). a…
74. WorkOS. Build secure ai agents, 2025. Accessed: 2025-05-13. [find](https://scholar.google.com/scholar?q=WorkOS%2E+Build+secure+ai+agents%2C+2025%2E+Accessed%3A+2025%2D05%2D13%2E)
75. Zhenliang Xue, Yixin Song, Zeyu Mi, Xinrui Zheng, Yubin Xia, and Haibo Chen. Powerinfer-2: Fast large language model inference on a smartphone. arXiv preprint arXiv:2406.06282, 2024. [find](https://scholar.google.com/scholar?q=Zhenliang+Xue%2C+Yixin+Song%2C+Zeyu+Mi%2C+Xinrui+Zheng%2C+Yubin+Xia%2C+and+Haibo+Chen%2E+Powerinfer%2D2%3A+Fast+large+language+model+inference+on+a+smartphone%2E+arXiv+preprint+arXiv%3A2406%2E06282%2C+2024%2E)
76. Biwei Yan, Kun Li, Minghui Xu, Yueyan Dong, Yue Zhang, Zhaochun Ren, and Xiuzhen Cheng. On protecting the data privacy of large language models (llms): A survey. arXiv preprint arXiv:2403.05156, 2024. — [arXiv:2403.05156](https://arxiv.org/abs/2403.05156) [find](https://scholar.google.com/scholar?q=Biwei+Yan%2C+Kun+Li%2C+Minghui+Xu%2C+Yueyan+Dong%2C+Yue+Zhang%2C+Zhaochun+Ren%2C+and+Xiuzhen+Cheng%2E+On+protecting+the+data+privacy+of+large+language+models+%28llms%29%3A+A+survey%2E+arXiv+preprint+arXiv%3A2403%2E05156%2C+2024%2E)
77. Fanjia Yan, Huanzhi Mao, Charlie Cheng-Jie Ji, Tianjun Zhang, Shishir G. Patil, Ion Stoica, and Joseph E. Gonzalez. Berkeley function calling leaderboard. [https://gorilla.cs.berkeley.edu/blogs/8_berkeley_function_calling_leaderboard.html](https://gorilla.cs.berkeley.edu/blogs/8_berkeley_function_calling_leaderboard.html), 2024. [↗](https://gorilla.cs.berkeley.edu/blogs/8_berkeley_function_calling_leaderboard.html)
78. Shunyu Yao, Noah Shinn, Pedram Razavi, and Karthik Narasimhan. Tau-bench: A benchmark for tool-agent-user interaction in real-world domains. arXiv preprint arXiv:2406.[12045, 2024](https://md.research.elamur.ai/free/en/papers/yao-2024-tau-bench-benchmark-tool/).
79. Da Yu, Peter Kairouz, Sewoong Oh, and Zheng Xu. Privacy-preserving instructions for aligning large language models. arXiv preprint arXiv:2402.13659, 2024. — [arXiv:2402.13659](https://arxiv.org/abs/2402.13659) [find](https://scholar.google.com/scholar?q=Da+Yu%2C+Peter+Kairouz%2C+Sewoong+Oh%2C+and+Zheng+Xu%2E+Privacy%2Dpreserving+instructions+for+aligning+large+language+models%2E+arXiv+preprint+arXiv%3A2402%2E13659%2C+2024%2E+%E2%80%94+%5BarXiv%3A2402%2E13659%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F24)
80. Adam Zewe. Like human brains, large language models reason about diverse data in a general way. MIT News, February 19 2025. Accessed: 2025-05-09. [find](https://scholar.google.com/scholar?q=Adam+Zewe%2E+Like+human+brains%2C+large+language+models+reason+about+diverse+data+in+a+general+way%2E+MIT+News%2C+February+19+2025%2E+Accessed%3A+2025%2D05%2D09%2E)
81. Jianguo Zhang, Tian Lan, Ming Zhu, Zuxin Liu, Thai Hoang, Shirley Kokane, Weiran Yao, Juntao Tan, Akshara Prabhakar, Haolin Chen, et al. [xlam: A family of large action models to empower ai agent systems](https://md.research.elamur.ai/free/en/papers/zhang-2025-xlam-family-large-action-models/). arXiv preprint arXiv:2409.03215, 2024.
82. Kevin Zhang. A deep dive on ai inference startups, 2024. Accessed: 2025-05-09. [find](https://scholar.google.com/scholar?q=Kevin+Zhang%2E+A+deep+dive+on+ai+inference+startups%2C+2024%2E+Accessed%3A+2025%2D05%2D09%2E)
83. Jeffrey Zhou, Tianjian Lu, Swaroop Mishra, Siddhartha Brahma, Sujoy Basu, Yi Luan, Denny Zhou, and Le Hou. [Instruction-following evaluation for large language models](https://md.research.elamur.ai/free/en/papers/zhou-2023-instruction-following-evaluation-large-language/). arXiv preprint arXiv:2311.07911, 2023.
84. Xuezhi Zhou, Nathanael Schärli, Yujie Hou, Jason Wei, Denny Zhou, Quoc V. Le, and Douwe Kiela. [Least-to-most prompting enables complex reasoning in large language models](https://md.research.elamur.ai/free/en/papers/al-2022-least-most-prompting-enables-complex/). arXiv preprint arXiv:2205.10625, 2022.
85. David Zier and Harry Kim. Introducing nvidia dynamo, a low-latency distributed inference framework for scaling reasoning ai models, March 2025. Accessed: 2025-05-09. [find](https://scholar.google.com/scholar?q=David+Zier+and+Harry+Kim%2E+Introducing+nvidia+dynamo%2C+a+low%2Dlatency+distributed+inference+framework+for+scaling+reasoning+ai+models%2C+March+2025%2E+Accessed%3A+2025%2D05%2D09%2E)

*Updated 2 November 2026*