News

ModelsNov 9, 20263 min read1 source

Meta releases cheap new AI model built for agent tasks

Worth watching

Meta says its new Ember model is far cheaper per use than rivals like Claude’s Haiku, aimed at the repetitive work AI agents do. It’s not yet clear how small businesses would access it.

Read it in

Drafted with AI· Nov 9, 2026

Models · Nov 9, 2026

The distilled version

  • Meta released Muse Ember, a lightweight AI model built for repetitive agent tasks like tool calls, coding subtasks and document parsing.
  • Meta says it costs 10 cents per million input tokens, cheaper than Anthropic’s Haiku and OpenAI’s Luna.
  • The sources don’t say when or how small businesses could use Ember directly, so there’s nothing to set up yet.

What happened

Meta Superintelligence Lab released Muse Ember on Monday. TechCrunch describes it as the lab’s lightest and most cost-efficient AI model yet. Meta says it is designed for the repetitive work AI agents do all day: tool calls, coding subtasks, routing, extraction and long document parsing.

Meta says Ember costs 10 cents per million input tokens, making it significantly cheaper than competitors like Anthropic’s Haiku and OpenAI’s Luna, and cost-competitive with open source models like DeepSeek V4.1 Flash and GLM 5.3 Flash. Meta also says Ember can handle text, images and audio, so agents could read screenshots, charts and scanned documents, or work from voice memos and recordings, without needing a separate model.

Meta suggests pairing Ember with a larger model like its own Muse Wonder, where the bigger model sets the rules and Ember applies them across large volumes of cases, such as a customer-support agent handling tens of thousands of cases. Ember is the fourth model released under Meta’s Muse brand, following Spark, Glimmer and Wonder.

Why it matters

If your business uses or is considering AI agents, for example to handle customer support tickets or sort documents at volume, the cost per task matters once you’re running thousands of them. A cheaper model built for that volume work could matter for anyone using Meta’s agent tools, though the sources don’t say what that setup costs in practice or who can access it.

What it means for your business

If you run the business: there’s nothing to change yet. This is Meta’s own infrastructure model, not a consumer app, so keep it on your radar rather than acting on it now.

What to do

Worth watching
  • Note the name Muse Ember and check back in a month for news on how businesses can access it.
  • If you already use an AI agent tool, ask the provider which model it runs on and whether cheaper options like Ember are planned.
  • Don’t switch anything yet: the sources don’t say Ember is available to small businesses directly.

Worth watching: no action needed yet; revisit when it matures.

Hype check

The claim: Meta says Ember is cheaper and smarter than competitors like Claude’s Haiku and OpenAI’s Luna.

The reality: That comparison comes from Meta’s own release using a third-party benchmark. The sources don’t say how it performs on real small business tasks, or when it would be available to use.

Jargon in this story

ModelAgentTokenMultimodalMini model

Sources

  1. 1Meta releases Muse Ember, a new lightweight model for agent execution (opens in a new tab) · TechCrunch · News report · Nov 9, 2026

Keep going: learn it, then do it.

04Builder14 min

Decide how you’ll judge it before you build it

Collect 20 real examples and what a good answer looks like for each.

Learn · 4 minTry it · 10 min
Start lesson
Agents & automation

What is an AI agent, really?

An AI that takes actions toward a goal, not just one that answers questions.

6 min