Meta Muse Wonder Guide

Muse Wonder is Meta’s newest family of AI models, launched on October 12, 2026. It is built for agent work: changing software, doing research, handling documents, and running multi-step tasks with tools.

It succeeds Muse Spark, the model behind the Muse personal agent. Like Muse Spark, and unlike Meta’s older Llama models, it is closed-weight.

Muse is the agent. Muse Wonder is the brain Meta is placing behind it.

Muse Wonder

The standard model, offered through the Meta Model API and Muse Code. Its model ID is muse-wonder-1.0.

Muse Wonder Full

Meta’s most capable option. At launch it is available only in Muse Code, on eligible Power and Maximum plans, with no public API price.

Muse Code

Meta’s coding agent app, out of beta since August 31, 2026

Meta Model API

For developers who build their own apps and agents

Muse app

Meta says Muse Wonder will power the personal Muse agent, reaching subscribers first

Rollouts can vary by account and country. Check what your own app or account shows.

Muse Wonder

$4 per million input tokens, $0.20 for cached input, $20 for output

Muse Spark 1.3 Standard

$1.25 input, $0.15 for cached input, $4.25 output

Muse Wonder costs roughly three to five times as much as Spark 1.3. Send simple, high-volume work to Spark and save Wonder for work where quality matters. Prices reflect launch, so check Meta’s pricing page.

Active Memory

Draws relevant context from your earlier conversations

Active Focus

Helps the model persist on hard tasks, adapt to changes, and ask you for input

Agentic reasoning

Plans and uses tools through long, multi-step tasks

On Meta’s own benchmark chart, Muse Wonder Full leads on some long coding benchmarks, like DeepSWE and SWE-Marathon, while Claude Opus 5.5 leads on others, like ProgramBench. Some comparisons relied on different agent setups.

No benchmark knows your work. Test on your own tasks before you switch.

• Download Muse Code from Meta’s official site only
• Begin in an empty practice folder
• Select Muse Wonder as the model and set the reasoning effort
• Begin with read-only permissions
• Check your settings, model, and usage allowance inside the app
• Hand it a small task with clear rules

• Pick three real tasks whose answers you already know
• Run them on Muse Wonder and on your current model
• Check facts, math, and code yourself
• Change a rule mid-task and watch whether it adapts
• Give it an impossible request and see whether it says so
• Compare quality, speed, and cost side by side

• Muse Code subscriptions default to Contributor mode, which lets Meta train on your prompts and responses. Switch it off in settings if you prefer
• Muse Code runs on your computer but sends the work to Meta’s servers
• Keep passwords, keys, and private client data out of prompts
• Test with non-sensitive material first
• Review exported transcripts before you share them

• Keep approvals on for file changes and commands
• Restrict network access for agent tasks
• Pressing Esc stops the current turn, but also check for background tasks and subagents
• Before long runs, confirm whether you are billed through a subscription or an API key

• Fixing bugs and building features in a codebase
• Research backed by sources
• Turning long documents into summaries and plans
• Multi-step agent workflows that use tools
• Planning tasks around clear rules and budgets

You are helping me with [task]. Rules: [fixed rules, limits, budget, deadlines]. Use only these files or sources: [list]. Work step by step, show your plan before you act, and cite a source for every fact. If you cannot meet a rule, stop and tell me rather than working around it. End with: what you did, what you are unsure of, and what needs my decision.