Free PDF/Technical Deep Dives
Model Routing
Sending Each Request to the Cheapest Model That Can Still Answer Correctly
A book by Alpesh Nakrani
Subscribe to get the complete book. Your PDF downloads immediately.
- Pages
- 174
- Chapters
- 19
- Format
- PDF · 30.5 MB
01/Overview
What's in the book
One model for everything is the expensive default. Routing requests to the cheapest model that can actually answer them.
02/Inside the PDF
Table of contents
PDF page- FMFront Matter: Model Routing—
- INTIntroduction: The Tower18
- 01The Bill That Broke the One-Model Religion24
- 02The Frontier, Not the Flagship32
- 03Static, Dynamic, Cascade, Fallback, Ensemble: Five Words That Are Not Synonyms39
- 04Prompt Length Is a Liar47
- 05Reading Risk Before Reading Difficulty54
- 06Confidence, Self-Assessment, and Why Models Lie About Both61
- 07Difficulty From History: Slices, Embeddings, and Learned Routers68
- 08Rules, Intent, and Tiers76
- 09The Cascade Ladder83
- 10Failover, Local-vs-Cloud, and the Provider Mesh90
- 11Ensembles, Voting, and Rerank-and-Fuse97
- 12The Confusion Matrix Has Four Boxes, Not Two104
- 13Regret, Oracles, and Cost-Weighted Quality111
- 14Shadow Routing and Online Evaluation118
- 15The Cost Waterfall125
- 16The Latency Budget132
- 17Locked Doors: Residency, Permissions, and Abuse138
- 18The Control Room and the Fleet146
- 19Ten Playbooks153
- AAppendix A: Back Matter162
The full chapters, illustrations, and reference material are included in the downloadable PDF.
Model Routing
The complete edition,
ready to read offline.
Alpesh's bookshelf
Get your free copy.
Subscribe to Alpesh’s book and writing updates. Unsubscribe anytime.
Enjoy your book
Your download is ready.
You’re subscribed. Your PDF download should start automatically.
Download PDF againIf it doesn’t start, use the button above.
