What Is Meta Muse Wonder? Price, Access and Full Version
The short answer
Muse Wonder is Meta’s frontier model for coding and agents, released October 12, 2026. Developers can call it on the Meta Model API as muse-wonder-1.0 for $4 input / $0.20 cached / $20 output per million tokens. A stronger variant, Muse Wonder Full, is only sold inside Meta’s Muse Code terminal agent on the two top subscription plans.
| Muse Wonder | Muse Wonder Full | |
|---|---|---|
| Released | October 12, 2026 | October 12, 2026 |
| Where | Meta Model API (Standard mode only) and every Muse Code plan | Muse Code Power Usage and Maximum Usage plans only |
| Price | $4 / $0.20 cached / $20 per 1M tokens; no long-context premium | Subscription only |
| Context | 1,048,576 tokens | — (Muse Code) |
| Inputs | Text, image, video, audio, PDF | Same family |
| AA Intelligence Index v4.3 (max) | 52 | 56 |
What it is built for
Meta’s model docs describe Muse Wonder as built for “complex software engineering and long-horizon agentic work”: making and verifying changes across many files in real repositories, planning, delegating to sub-agents, calling tools in parallel and recovering from errors. It also does computer use, structured output, tool calling and search grounding ($2.50 per 1,000 searches on top of tokens). It replaces Muse Spark 1.3 (September 2, 2026) at the top of Meta’s line-up; Muse Spark stays on sale at $1.25 / $4.25.
Meta’s benchmark chart covers agentic and professional tasks — Harvey legal agent work, Vals Finance Agent, HealthBench Pro, AutomationBench, OSWorld 2.1, Terminal-Bench 4.0, DeepSWE 1.1, SWE-Marathon and others. Those are Meta’s own numbers; the independent figure is Artificial Analysis: 52 for Muse Wonder and 56 for Muse Wonder Full at max effort, a material jump from Muse Spark 1.3’s 48 six weeks earlier.
How to get it
- API: create a Meta Model API key and call
muse-wonder-1.0through the Responses, Chat Completions or Messages endpoints. Contributor mode (cheap tokens in exchange for training on your data) is not available for Muse Wonder on the API. - Muse Code subscription: four plans — Everyday Usage, High Usage (5x), Power Usage (20x, adds Muse Wonder Full) and Maximum Usage (80x). Subscriptions default to Contributor mode, which you can turn off.
How it compares on price
| Model | Input / output per 1M | AA Index (current version) |
|---|---|---|
| Claude Opus 5.5 | $4 / $20 | 58 |
| Meta Muse Wonder | $4 / $20 | 52 |
| GPT-6.1 Sol | $2 / $10 | 52 |
| Gemini 4 Argon | $2 / $10 (intro to Nov 30, 2026; $4 / $20 from Dec 1) | 53 |
| Muse Spark 1.3 | $1.25 / $4.25 | 48 |
On list price and the independent index, Muse Wonder is priced like Opus 5.5 and scores like GPT-6.1 Sol. It makes sense if you already build on Meta’s API or Muse Code, or if your own evals favour it on long agentic coding runs. For the full price table see current API prices; for Meta’s open-weight line see what is Muse Spark Open.
Last verified: October 18, 2026.