What's inside#
- Chapter 1 — The Big Bet: the evidence that small, specialized models beat giants — Orca-2-13B nearly catching GPT-3.5, Meerkat-7B topping a 70B medical model
- Chapters 2–4 — The wins: medicine, law, and finance; code, math, and reasoning; language, retrieval, and speech
- Chapters 5–6 — The mechanism: how fine-tuning actually works, and why data beats scale
- Chapter 7 — The economics: the full invoice math — ~$3,200/month vs ~$96/month, a 33× gap, and the break-even analysis
- Chapters 8–9 — The honesty: when fine-tuning fails, and where scale still wins
- Chapter 10 — The Playbook: NVIDIA's published recipe for replacing giant models with small ones, plus a migration checklist
Plus a glossary. Written in-between technical and non-technical: precise enough for engineers, clear enough for the people who sign the API bills.
Who it's for#
- Engineering leaders and ML practitioners deciding what to fine-tune and what to leave alone
- Founders and operators paying a growing AI API bill
- Anyone who suspects the default “just use the biggest model” advice is costing them money
Download the full book — free#
Small Models Win is free. All 10 chapters plus the glossary, in both formats:
Prefer a taste first? The free sample PDF (preface + chapters 1–2) is also available.
FAQ#
Do I need to be an ML engineer to read this?
No. The book is written in-between technical and non-technical: the evidence is precise, the explanations assume no ML background.
Are the benchmark numbers real?
Yes — every figure is sourced from published papers and official reports, with per-chapter sources. Prices are September 2026 snapshots and labeled as such.
Does the book admit where small models lose?
Two full chapters: when fine-tuning fails, and where scale still wins. The thesis is conditional, not a slogan.
What if prices or models change?
Ebook buyers get free lifetime updates.
Can I share it with my team?
The Team tier covers up to 10 people.
Read the first two chapters free#
Preface + chapters 1–2 of Small Models Win — the big bet, and the specialists at work in medicine, law, and finance. Free PDF, no spam.
Download the free sample (PDF)
Also by Yahya Laraki: Recommender Systems: From Ratings to Retrieval — a beginner-friendly, 142-page guide with runnable Python in every chapter.