8 Types of LLMs Powering AI Agents, from GPT Backbones to LAM Action Models
mdancho84 · x · 2026-11-07
Matt Dancho breaks down the 8 types of LLMs used in AI agent systems:
- GPT: general-purpose text understanding and generation backbone
- MoE: routes tokens to specialized "experts" to scale capacity efficiently
- LRM: tuned for multi-step reasoning plus RAG/tool use
- VLM: processes images + text for multimodal perception and grounding
- SLM: compact, fast models for edge/on-device or low-latency tasks
- LAM: plans and executes actions via tools/APIs/robots — the agent's actuation layer
- HLM: layered submodels (user/item/task) coordinating complex workflows
- LCM: maps words to higher-level concepts for abstraction and semantic understanding
A beginner-friendly primer on how model roles are layered inside an agent stack.
More from coding & agent
- PSA: AI agents love slapping broad :has() rules on CSS, causing huge style recalcs — jh3yy · 2026-11-07
- YC-backed Trepo debuts kitchen hardware that lets AI agents shop for groceries in the background — ycombinator · 2026-11-07
- Cognition ships Devin Review 2.0: 5× cheaper, 1.5× faster PR review — smtabatabaie · 2026-11-07
- Air moves its agentic development experience directly into JetBrains IDEs, free tiers included — therealdanvega · 2026-11-07
- PSA: keep CSS :has() rules tight — agents' broad selectors trigger huge style recalcs — jh3yy · 2026-11-07
- Solo founder credits AI chief-of-staff bots for record-breaking revenue month — joshgonsalves_ · 2026-11-07