Meta launches Muse Wonder and Wonder Full, betting on long-term Active Memory
kman583 · reddit · 2026-10-13
A Reddit post reports Meta released Muse Wonder and Wonder Full, scoring just behind Claude Opus 5.5 and billed as Meta's first genuine frontier model.
The poster argues the headline feature isn't the benchmark but two capabilities:
- Active Memory: the model carries what it knows about you across conversations over days and weeks
- Active Focus: it keeps grinding on a hard goal instead of quitting at the first blocker
Their take: in assistants, the model-IQ race is noise—the winner remembers your life and works while you're not watching. Wonder ships to paying subscribers first in the Muse app (invite code in post grants both sides 1B tokens).
Related event: Meta Unveils Frontier Model Muse Wonder for Personal AI Agents(102 posts)→
More from Models
- DeepSeek v4.1 self-review: solver correct to 14 digits, wrong problem — teortaxesTex · 2026-10-13
- Announcing AA-3D: Artificial Analysis benchmarks AI models building interactive 3D scenes — ArtificialAnlys · 2026-10-13
- Same models top almost every AA-3D domain, Artificial Analysis finds — ArtificialAnlys · 2026-10-13
- Scale's Alexandr Wang says muse-wonder bug fixes are already showing clear improvement — alexandr_wang · 2026-10-13
- Muse's $15 plan tested: roughly 1B tokens a month, ~$280 at API rates — Aldarund · 2026-10-13
- Fine-tuning speech models for Algerian Darija: a team's four open questions — Fair_Bed_914 · 2026-10-13