Models / Meta / Muse Wonder

Meta · muse

Muse Wonder

Muse Wonder is a frontier multimodal reasoning model from Meta for personal and professional agents, long-horizon coding, and computer use. It combines Active Memory, which carries relevant context across conversations, with Active Focus, which keeps it working through blockers toward a goal.

$4.00 in · $20.00 out

per 1M tokens · first-party price. Reference price from Meta, the lab's own listing.

Context
1,048,576
Max output
131,072
Cache read
$0.20
Cache write
—
Reasoning
Yes
Tool calls
Yes
Structured output
Yes
Vision
Yes
Open weights
No
License
—
Released
Oct 12, 2026
Updated in source
Oct 12, 2026
Knowledge
—
Providers
6

Announcement · Model card

What this model costs per month

Prices are USD per 1M tokens. A month is 30 days. Cache hits use the cache-read rate when the model has one; otherwise the cache field is ignored.

monthly = requests/day × 30 × (input tokens × (1 − cache hit) × input price + input tokens × cache hit × cache-read price + output tokens × output price) ÷ 1,000,000

Providers

Base rates from Models.dev. One row per provider.
ProviderInputOutputCache readCache writeStatus
Cloudflare AI Gateway$4.00$20.00$0.20——
DevPass$4.00$20.00$0.20——
LLM Gateway$4.00$20.00$0.20——
Metareference$4.00$20.00$0.20——
Requesty$4.00$20.00$0.20——
Vercel AI Gateway$4.00$20.00$0.20——

Jobs this model qualifies for