AI Daily Grind // AI news, freshly ground daily

← All lessons

🧠 Models and how they work · ⏱️ 2 min Latte

What is Small model?

Illustration: Small model
Simply put, like I'm 5 years oldA small model is like a little kitchen knife next to a big chef's knife. It cannot do every job, but for cutting a lot of small things fast, it is lighter, quicker and cheaper to keep.

📖 The slightly longer version

An AI model learns by adjusting millions or billions of internal numbers, called parameters. A small model has fewer of them. That makes it lighter: it needs less computer memory and less work from the chips for each answer. Less work means faster answers and a lower price per token (a token is a small piece of text, the unit companies use to count and charge).

The trade-off is simple. A big model is usually better at hard thinking, long reasoning and rare questions. A small model is good at simple, repeated jobs: sorting text, pulling a fact out of a document, calling a tool. Many companies now pair them: a big model plans, and small models do the many little steps. Meta says its new Muse Ember is its smallest and most efficient model, built for this kind of agent work.

Why a small model costs less

  1. Fewer parametersThe model has fewer internal numbers to adjust.
  2. Less computingEach answer needs less memory and less chip work.
  3. Lower token priceThe company can charge less per piece of text.
  4. Good for repeat jobsAgents use it for many simple steps, like calling tools.
  5. More agent actionsCheaper steps mean more actions, and more chances for errors.
Why it matters

Agents repeat small jobs all day, and every step costs tokens. Meta prices Muse Ember at $0.10 per million input tokens and $0.50 per million output tokens for shorter prompts, so running an agent gets much cheaper. Cheaper agents mean more of them, doing more actions in our files, inboxes and accounts. Useful, yes, and also many more chances to touch the wrong key.

⚠️ What to watch out for

"Efficient" and "much cheaper" are mostly claims from the company and early reports, not long independent tests. Simon Willison notes that for shorter prompts Ember costs the same as GPT-6 Luna and Claude Haiku 5.5, so "cheaper than rivals" depends on how you use it. A small model can also make more mistakes on tricky tasks, and Ember is closed, so outsiders cannot inspect it.

📰 Where it showed up

This notion came up in A new Claude takes the top spot while trust cracks everywhere else. More lessons: Learn.

🎯 Keep learning

New 🧠 Models and how they work · 2 min Latte What is Quantization? It is like a recipe that says 'about one cup of flour' instead of '236.6 grams'. The recipe card gets much shorter, and the cake tastes almost the same. Learn it → 🧠 Models and how they work · 2 min Latte What is Chain-of-thought? It is like a cook who says each step out loud while making soup: "now the onions, now the salt." If you listen, you can stop them before they add something bad. Learn it → 🧠 Models and how they work · 2 min Latte What is Token? A token is like one small piece of a chopped carrot. The AI does not eat a whole sentence at once: it cuts the words into little pieces and reads them one by one. Learn it →

Sources

Written on 2026-11-10. This explanation was accurate on that day. AI changes fast, so check the sources and the date before you rely on it.

Get The AI Daily Grind by email, every morning, in simple words.