---
id: "prereq-llm-architectures"
type: "prereq"
source_timestamps: ["§ What We Found"]
tags: ["ai-literacy"]
related: ["concept-reasoning-vs-non-reasoning-models"]
reason: "Required to grasp why different models react differently to the same promotional cues."
sources: ["geo"]
sourceVaultSlug: "hbr-seg-geo"
originDay: 3
articleStem: "hbr-tier2-06-ai-shopping-agents"
sourceUrl: "https://hbr.org/2026/05/research-traditional-marketing-doesnt-work-on-ai-shopping-agents"
sourceTitle: "Research: Traditional Marketing Doesn’t Work on AI Shopping Agents"
---
# LLM Capabilities and Architectures

**Prerequisite knowledge:** Basic familiarity with the LLM landscape — specifically the distinction between **"reasoning" models** (e.g., [[entity-gpt-5|GPT-5]], [[entity-gemini-2-5-pro|Gemini 2.5 Pro]]) and **"non-reasoning" / lighter models** (e.g., [[entity-gemini-2-5-flash-lite|Flash Lite]], [[entity-gpt-4-1-mini|GPT-4.1-mini]]).

**Why it's required:** Without this distinction, the central empirical finding is unreadable — you need it to grasp *why* different models react differently to the same promotional cue, and why [[concept-reasoning-vs-non-reasoning-models|the reasoning vs. non-reasoning split]] underwrites [[concept-ai-model-segmentation|model segmentation]].

**Helpful context:** Reasoning-heavy models are multimodal, tool-enabled systems with explicit "Thinking" / "Deep Think" modes that scrutinize information more deeply; lighter models are optimized for speed and cost.

**Related:** [[concept-reasoning-vs-non-reasoning-models]] · [[concept-ai-model-segmentation]] · [[concept-algorithmic-skepticism]]