Modern AI Engineering

Module 7 · Core

The Language Model Zoo

In this module, we will learn that not every language model is a large, text-generating LLM. We will see the smaller, reasoning, recursive, diffusion-based, and decision-only models and when to use which one.

By the end of this module, we will know which type of model to pick for a given problem.

Lessons

  1. 7.1 Small Language Models: Big Capability in Compact Form: SLM = Small + Language Model · What is a Language Model? · What Counts as "Small"? · Popular SLMs we should know · How SLMs Stay Capable Despite Being Small · Why SLMs Matter · SLM vs LLM · The Size Spectrum · Where SLMs Shine - Use Cases · Trade-offs of SLMs · When to Pick an SLM · Quick Summary
  2. 7.2 Large Reasoning Models: Chain-of-Thought at Inference Time: The Big Picture · What is a Large Reasoning Model (LRM)? · LLM vs LRM · How does an LRM actually think? · Test-time compute: thinking longer makes them smarter · How are LRMs trained? · Input and Output: training phase vs prediction phase · When to use an LRM, and when to use a regular LLM · Popular LRMs we should know · Common Mistakes when using LRMs · Quick Summary
  3. 7.3 Recursive Language Models: Self-Referential Generation: What is a Recursive Language Model (RLM)? · Why do we need RLMs? · How an RLM works · How the model writes and runs code · Why RLMs work better · Recursion inside RLMs · How RLMs differ from simple chunking · Advantages of RLMs · Limitations of RLMs · When to use RLMs · RLM vs RAG · A real use case
  4. 7.4 Diffusion Language Models: Text Generation Beyond Autoregression: What is a Diffusion Language Model? · How do today's language models write text? · The problem with the usual approach · Where the diffusion idea comes from · What does "noise" mean for text? · The two phases: forward and reverse · How a DLM actually generates text, step by step · A tiny end-to-end example · A simple code-style walk-through · DLMs vs the usual language models · Advantages of DLMs · Limitations of DLMs · The current state
  5. 7.5 Jev and System One: Fast vs Deliberate AI Thinking: What is a System One Model? · System One vs System Two Thinking · The Problem with Using an LLM for Decisions · What is Jev? · How Jev Works · Typed Answers: Choice, Score, and Yes/No · Calibration and RLCD · Why Jev Cannot Hallucinate · Jev vs LLM · Where Jev Works Well and Where It Fails · When to Use Which One

← Module 6: Next-Gen LLM Architectures · Module 8: Teaching and Shaping Models →