No. 470 of 500 No. 62 in Generative AI & LLMs

Domain-Specific Small Language Models

Efficient AI for local deployment

By Guglielmo Iozzia

Published
May 26, 2026
Print length
376 pages
Reading level
Advanced
ISBN-13
978-1633436701
Publisher
Manning Publications
Written for
AI engineers and ML practitioners deploying specialized language models under cost, privacy or hardware constraints.

Synopsis

Guglielmo Iozzia argues that many organizations do not need a giant general-purpose model to answer questions in a narrow field, and that a smaller model adapted to domain data can be faster, cheaper, easier to keep private and simpler to run within regulatory limits. The book explains what distinguishes small language models, how encoder and decoder architectures differ, and how to size a model for a task. It then works through adapting pretrained open-source models with transfer learning and parameter-efficient fine-tuning on custom datasets, using Hugging Face libraries and other open-source frameworks, runtimes and utilities.

A large share of the book addresses deployment: quantization and other optimizations that shrink memory footprint, serving models through secure APIs, and running them on commodity machines and edge devices. Examples span general software tasks, such as Python code generation, and scientific ones that reflect the author's biomedical specialty, including protein structure and antibody sequence generation. The book also treats small models as components of agentic systems. Readers finish with a workflow for taking an open model from selection through fine-tuning to an optimized local deployment, along with the judgment to decide when a small specialist model is a better choice than a hosted frontier model.

Scorecard

Editorial score

How it scored

Editorial score: 65.5 / 1005 weighted criteria

  • Influence30%

    12.0/30
  • Authority25%

    17.5/25
  • Practical value20%

    16.0/20
  • Currency15%

    13.5/15
  • Readability10%

    6.5/10
  • Editorial score

    65.5/100

How the rank was computed

Rank No. 470 of 500 comes from a composite of 48.0: observable signals 18.5 of 55 points plus editorial review 29.5 of 45 points.

All 11 inputs
InputPoints
Observable signals18.5/55
Reach (Amazon ratings volume)5.7/16.5
Consensus across best-of lists2.8/13.8
Awards and bestseller evidence0.0/8.3
Reader rating, sample-adjusted6.2/8.3
Amazon sales rank1.0/5.5
Recency2.8/2.8
Editorial review29.5/45
Author authority7.9/11.3
Influence5.4/13.5
Practical value7.2/9.0
Readability2.9/4.5
Currency6.1/6.8

Each criterion is scored from 1 to 10 and weighted, and the points add up to the editorial score. The rank combines observable signals (55 percent) with this score (45 percent). Read the methodology.

Verdict

A timely, hands-on guide to fine-tuning, quantizing and running small specialized language models on modest hardware, written by an industry practitioner.

Strengths

  • Covers the full path from fine-tuning to deployment
  • Unusual scientific examples in protein and antibody design
  • Reflects current open-source tooling as of 2026

Keep in mind

Readers need solid Python and some machine learning background, and the fast-changing open-model ecosystem will date specific library versions and model choices.

Key takeaways

  1. For narrow tasks, a small model fine-tuned on domain data can rival much larger general models in accuracy and speed at lower cost.
  2. Parameter-efficient fine-tuning adapts pretrained open-source models to new domains without the compute budget that full retraining requires.
  3. Quantization and related optimizations reduce memory requirements enough to run capable models on commodity hardware and edge devices.
  4. Local deployment keeps sensitive data in-house, which helps organizations meet privacy and regulatory requirements in fields such as healthcare.

Notable facts

About the author

Guglielmo Iozzia

Guglielmo Iozzia is Director of AI and Applied Mathematics at Merck & Co., known as MSD outside the United States, where he specializes in biomedical AI applications. Manning describes his industry experience across biotech manufacturing, healthcare, cloud operations and cybersecurity, and he is a Distinguished Member of the American Society for Artificial Intelligence.

Questions

Is Domain-Specific Small Language Models worth reading in 2026?
Yes, if you build or deploy language models. Published in 2026, it reflects current interest in smaller open models that run locally for cost, latency and privacy reasons. It is most valuable for teams in regulated or hardware-constrained settings, such as healthcare, finance or edge devices, that need specialist accuracy without sending data to a large hosted model.
Who should read Domain-Specific Small Language Models?
AI engineers, machine learning practitioners and technical leads who know Python and want to adapt open-source models to a specialized field. It assumes comfort with code and basic deep learning concepts. Managers looking for a nontechnical overview of small language models, and readers new to programming, should start with a more introductory book.
How do small language models differ from large ones?
In Iozzia's framing, a small language model has far fewer parameters and lower memory and compute needs than a frontier LLM, so it can run on commodity hardware or edge devices. Fine-tuned on domain-specific data, it can match larger general models on focused tasks while offering lower cost, faster responses and stronger data privacy.

Sources

As an Amazon Associate, Best AI Books earns from qualifying purchases. Best AI Books is sponsored by Iternal.ai, whose founder and CEO, John Byron Hanby IV, wrote The AI Strategy Blueprint, our #1 ranked book. How we rank books.