Meta has released Muse Ember, the smallest model in its Muse family, and the number that matters is the price: roughly 10 cents per million input tokens, with a one-million-token context window. In a market where the flagship models cost dollars per million and arrive wrapped in talk of superintelligence, a capable model that costs about the same as a packet of crisps is its own kind of statement.
The giant-killer pitch
“Giant killer” is a label the industry reaches for every few months, usually prematurely. The honest version here is narrower: Muse Ember is not going to top the frontier leaderboards, and it is not meant to. It is built to be cheap, fast and good enough for the enormous category of tasks that do not need a genius, only a competent, tireless worker. Summarising, tagging, routing, drafting, extracting: the unglamorous bulk of real AI work, done at a tenth of the cost.
Why Meta, of all people
There is a strategic read here that goes beyond one cheap model. Scott Galloway, never shy of a market call, recently named Meta the company he expects to look back on as the real winner of the AI boom. His argument is not that Meta has the best model. It is that Meta has a gigantic, cash-generating advertising business funding the whole adventure, and the capital to keep fighting the arms race long after thinner players run out of road. While OpenAI and Anthropic absorb most of the capex and most of the risk, Meta can treat AI as a feature of a business that already works, and sell shovels, including leasing out compute, rather than betting the company on the gold.
Galloway has paired that with a blunter warning: that an unusual share of the entire AI build-out now rests on demand from only two buyers, and that the whole thing rhymes with 1999. If he is right, the company positioned to survive the hangover is the one that was never dependent on the party continuing. A 10-cent model is exactly what that strategy looks like from the outside.
What this means
If you are building anything on AI, Muse Ember is worth a look for the high-volume, low-stakes jobs where paying flagship prices is simply waste. Do not expect brilliance; expect cheapness and scale, which for most workloads is the better deal. And if Galloway’s thesis holds, the lesson is bigger than one model: in a bubble, the winner is often the one selling the cheap, boring, indispensable thing while everyone else chases the miracle.
Not investment advice. Sources: model-release coverage, November 2026; Scott Galloway, Prof G Markets.