# Meta Active Memory and Active Focus — Muse Wonder trains the Muse app models

> Meta uses its top model, Muse Wonder, to train the everyday models in Muse and Meta AI to remember more and give up less.

- Org: Meta
- Date: 2026-11-06
- Category: model
- Source: https://research.meta.ai/blog/specialized-models-for-muse-and-metaai
- Canonical page: https://ai-tldr.dev/releases/meta-muse-active-memory-active-focus/

## In short

Meta is rolling out specialized versions of the models behind Muse and Meta AI, trained using Muse Wonder. They add Active Memory and Active Focus, and Muse now scores 47% on Meta's Personal Superintelligence Index, up from 38%.

## What it is

Active Memory and Active Focus are two new abilities in the Muse and Meta AI assistants. Meta delivers them through specialized versions of the models behind both products, trained using Muse Wonder, which Meta introduced a few weeks earlier. This is not a rollout of the Muse Wonder models themselves.

## How it works

For Active Memory, Meta trained the models to tell temporary information from permanent information, to reason across time, and to spot details that have gone stale. For Active Focus, training covered real-world bottlenecks, so the assistant can try harder, recover, or find another route when a task hits an unexpected problem.

## Why it matters

Most people use Meta's assistants, not the expensive Muse Wonder API. Meta reports a 2.4x drop in forgetting critical information and a 67% improvement on everyday tasks where unexpected issues come up, so the frontier model's gains now reach the free, mass-market products.

## Who it's for

Muse and Meta AI users

## Try it

Open Muse or Meta AI — the update is rolling out to all users

## Quick facts

- Maker: Meta
- Where: Muse and Meta AI assistants
- Trained with: Muse Wonder
- What's new: Active Memory + Active Focus
- Active Memory Eval: 90% (was 76%)
- Pricing / API: Not announced

## FAQ

### Is Meta giving Muse Wonder to all Muse and Meta AI users?

No. Meta is rolling out specialized versions of the models already behind Muse and Meta AI, trained using Muse Wonder. Unite.AI notes this is not a rollout of the Wonder series itself. Muse Wonder stays Meta's most capable model, and here it acts as the teacher that improves the everyday assistant models.

### How much better is the new Muse model than the previous one?

Meta reports that the current Muse model scores 47% on its Personal Superintelligence Index, up from 38% for the previous model. On Meta's Active Memory Eval it correctly uses critical context in 90% of tasks, versus 76% before. Meta also reports 2.4x less forgetting and a 67% gain on everyday tasks with unexpected issues.

### Can developers use the Active Memory models through the Meta API?

Meta's announcement for Active Memory and Active Focus names only the Muse and Meta AI products. It does not mention an API, pricing, or a model name for developers, so for now the new specialized assistant models are described only as updates inside Meta's own apps.

### What does Active Focus change in Meta AI?

Active Focus keeps Muse and Meta AI working toward a goal when something goes wrong mid-task. According to Unite.AI's report on Meta's post, Meta AI should split complex problems into steps, ask sharper follow-up questions, and adapt when users change direction, instead of stopping at the first obstacle.

## Links

- [Meta Announcement](https://research.meta.ai/blog/specialized-models-for-muse-and-metaai)
- [The New Stack](https://thenewstack.io/meta-muse-specialized-models/)
- [Unite.AI Report](https://www.unite.ai/current-muse-scores-47-on-metas-personal-superintelligence-index/)
- [Muse Wonder Launch](https://research.meta.ai/blog/muse-wonder)