SUN, NOVEMBER 08, 2026
Independent · In‑Depth · Practitioner‑Tested
✎ LLMs

Meta’s AI Now Remembers You, and Builds Rooms Around What You Do

Meta announced two things on 6 November. New specialised models bring Active Memory and Active Focus to every Muse and Meta AI user. Separately, Spaces lets you build persistent custom areas by describing them, with connected apps updating them in the background. Muse Wonder and Wonder Full stay paid.

By AIToolsRecap November 7, 2026 6 min read 18 views
Home › Articles › LLMs › Meta Muse Gets Memory and Custom Spaces, Free
The short answer
● Announced — 6 November, two separate features on the same day
● Active Memory — carries details across conversations, flags stale facts
● Active Focus — keeps an agent going when a task hits a problem
● Spaces — persistent custom rooms you build by describing them
● Who pays — the new models are for everyone, Muse Wonder stays paid

Meta shipped two things on 6 November that are being reported as one. They are separate, and the distinction matters if you are deciding whether to pay.

Active Memory and Active Focus

New specialised models, trained using Muse Wonder, roll out to every Muse and Meta AI user. Active Memory carries important details between conversations and is built to tell temporary information from lasting information, flag facts that have gone stale, and surface earlier details when they become relevant again.

Active Focus is the agentic half: when a task hits an unexpected problem, the model tries to recover or route around it rather than stopping and handing the problem back.

Spaces

A Space is a saved, persistent area inside the Muse app that you create by describing what you want, then refine by asking for changes. Meta gives examples including a holiday gift tracker, a daily word game and a DIY planning board.

Connected apps let a Space update in the background, so a concert date or a price can change without you asking. You can pin Spaces to the sidebar or main tabs, hide them, or schedule when they appear.

Sharing and collaborative Spaces are still being explored, so the first release is single-player.

What this does not change

Thing Status
New specialised models All Muse and Meta AI users
Spaces Everywhere Muse is available
Muse Wonder Still paid, via the Meta Model API
Muse Wonder Full Still paid, in the $50 and $200 Muse Code plans

Muse Wonder and Muse Wonder Full do not move to the free tier. What everyone gets is the smaller specialised models trained using Wonder, which is a different thing.

What Meta has not said

  • Which regions the new models are live in.
  • Any rollout schedule, so some users will see this later than others.
  • The names or sizes of the new specialised models, or how their training differs from the Muse Spark models used at the September launch.
  • Which plans or user groups get Spaces, beyond "wherever Muse is available".

Both memory benchmarks are Meta’s own and no third-party evaluation has been published. On Meta’s own figures the assistant still misses relevant context in roughly one task in ten.

Frequently asked questions

Do I have to pay for Active Memory?

No. The new specialised models carrying Active Memory and Active Focus go to all Muse and Meta AI users. Muse Wonder and Wonder Full remain paid.

What is a Space?

A persistent custom area in the Muse app that you create by describing it, which can pull live updates from connected apps.

Can I share a Space with someone?

Not yet. Meta says sharing and collaboration are still under exploration.

How good is the memory?

On Meta's own benchmark, it still misses relevant context in around one task in ten. No independent evaluation has been published.

Is this available in my country?

Meta has not stated which regions the new models are live in, or given a rollout schedule.

Related reading

Sources: Data Studios · Unite.ai.

Published 7 November 2026. Figures checked against the reporting linked above on the same day.

Tags
AI News2026AI agentsProductivity
⚑

Spot an inaccuracy?

We verify facts before publishing and correct errors promptly. If something in this article is wrong or outdated, let us know.

Report an error →