EUVIMEDEUVIMED entdecken
Crossref

Beyond Large Language Models: How Small, Specialized Agents are Replacing Monolithic Systems

Pritha Nandi, Subhajit Manna

International Journal of Science, Strategic Management and Technology · 29.03.2026 · 02 (03) · 1-9

DOI: 10.55041/ijsmt.v2i3.380 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Science, Strategic Management and Technology, online erschienen am 29.03.2026. Verfasst von Pritha Nandi und 1 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

. For years, the AI field kept chasing the bigger models where more parameters, more compute and more cost was observed. But by 2024, teams actually deploying AI in hospitals, factories, and satellite systems started hitting a wall where large models were too slow, too expensive, and too risky for private data. This paper takes a look at the changes that have happened, since then. Numerous researchers have splitted AI into smaller, focused parts that work together. Here, instead of one expert doing everything, the focus is on a team where each person handles what they do best. It was observed that three things matter most. First, separating planning, memory, and action into distinct modules makes systems far easier to fix and upgrade. Second, mixing small models for fast edge tasks with larger models for complex reasoning cuts costs by 40-70% without loss of output quality. Third, giving agents structured ways to communicate and disagree actually improves results. Controlled disagreement turns out to be useful. It was noted that two coordination patterns gain real traction: Mixture-of-Agents, where multiple models vote on the best answer, and Communicative Agents for "Mind" Exploration of Large Language Model Society (CAMEL) [1], where agents negotiate through role-based dialogue. Both patterns appeared to be promising .But neither is production-ready without human oversight. Thus, it can be said that there is no agreed standard for how agents across different systems should talk to each other. Until that exists, reproducing results from one lab to another stays harder than it should be. This is the most important unsolved problem in this area right now.

Zitieren

APA

Nandi, P., & Manna, S. (2026). Beyond Large Language Models: How Small, Specialized Agents are Replacing Monolithic Systems. International Journal of Science, Strategic Management and Technology, 02(03), 1-9. https://doi.org/10.55041/ijsmt.v2i3.380

Vancouver

Nandi P, Manna S. Beyond Large Language Models: How Small, Specialized Agents are Replacing Monolithic Systems. International Journal of Science, Strategic Management and Technology. 2026;02(03):1-9. doi:10.55041/ijsmt.v2i3.380

BibTeX
@article{Nandi2026,
  author = {Nandi, Pritha and Manna, Subhajit},
  title = {Beyond Large Language Models: How Small, Specialized Agents are Replacing Monolithic Systems},
  journal = {International Journal of Science, Strategic Management and Technology},
  year = {2026},
  volume = {02},
  number = {03},
  pages = {1--9},
  doi = {10.55041/ijsmt.v2i3.380}
}
Kostenlos starten →