Meta launches Muse Wonder and Muse Wonder Full: frontier agent models with Active Memory and Active Focus; API and Muse Code today, Muse assistant for subscribers next
0 reads
- Published
- 2026-10-13
- Type
- Model release
- Organization
- Meta
- Source
- agipulse
Why it matters
【Key points】
Meta on 2026-10-12 launched Muse Wonder and Muse Wonder Full, calling them its most capable models for personal and professional agentic work (claims, leases, finances, shipping apps). Newsroom ~17:08Z; @AIatMeta ~17:02Z, with Zuckerberg and Alexandr Wang same day. Coverage confirms prior codename Watermelon.
Access: Muse Wonder is live via the Meta Model API and Muse Code; Muse Wonder Full is for Muse Code subscribers. Meta says Wonder will soon power the Muse assistant, subscribers first. Training ran at the gigawatt-scale Prometheus data center after Muse Spark.
Pitch: Active Memory (carry relevant context across chats) and Active Focus (persist on hard multi-step work, recover, confirm before important actions). Alignment, private-data discretion, refusal of truly harmful asks, and cyber defenses are emphasized; a fuller safety report is “coming soon.” Frontier benchmark claims are vendor/press—no independent replication. No open weights announced.
【Verifiable】
Meta Newsroom: https://about.fb.com/news/2026/10/introducing-muse-wonder-models/
Meta AI Research: https://research.meta.ai/blog/muse-wonder
@AIatMeta (17:02Z): https://x.com/AIatMeta/status/2109691148943999167
@finkd / @alexandr_wang same-day. HF meta-llama: no matching new repo this patrol.
【Our read】
1. First clear “frontier” flagship ~14 months into Meta Superintelligence Labs’ rebuild: Spark → Wonder shifts from stack rebuild to productizing “personal superintelligence,” via API/Code tiers—not another open-weight drop.
2. Active Memory / Active Focus target assistant retention and long-horizon tasks—the Muse product gap that matters more than launch adjectives. Watch whether Wonder reverses Muse adoption softness.
3. Watermelon hype flipped into a multi-thousand-post/hour spike, but evals remain seller-stated. Next: model card, external benches, and Muse subscriber rollout pace.
Related developments
Aggregated by AGI Pulse. Titles and links only; no full text is republished. Editorial policy