What is Muse Wonder 1.0?
Muse Wonder 1.0 is Meta's frontier model for personal and professional agents. Released October 12, 2026 alongside Muse Wonder Full, it is available through the Meta Model API and Muse Code. Wonder Full is the higher-end subscription-only variant, while muse-wonder-1.0 is the model developers can call directly by API.
Meta describes two central capabilities: Active Memory, which carries relevant information across conversations, and Active Focus, which keeps working through blockers and changing information. The model is designed to ask for clarification and check in before important actions rather than acting as an unrestricted autonomous system.
QUICK VERDICT: Muse Wonder 1.0 is Meta's API-accessible frontier agent model, with a one-million-token context window and native text, image, video, audio and PDF input. Its Active Memory and Active Focus features are designed for persistent work, but that same access to personal context makes permission design, privacy and approval checkpoints central deployment requirements.
Field | Verified value |
|---|---|
Provider | Meta |
Release date | October 12, 2026 |
License | Proprietary API model |
API model ID | muse-wonder-1.0 |
Context window | 1,048,576 tokens |
Input modalities | Text, image, video, audio and PDF |
Output modality | Text |
Reasoning effort | Minimal through max |
Primary access | Meta Model API and Muse Code |
Standard API price | $4 input, $0.20 cached input and $20 output per million tokens |
Long-context premium | None documented by Meta |
Last verified | October 22, 2026 |
What changed from Muse Spark 1.3
Muse Wonder is a larger frontier step beyond Muse Spark rather than a routine speed or price update. It builds on a stronger base model and adds more emphasis on persistent memory, sustained execution, domain expertise and long-running coding projects.
Area | Muse Wonder 1.0 change |
|---|---|
Model tier | Moves from the Spark line to Meta's frontier Wonder series |
Memory | Active Memory retrieves earlier personal and task context when relevant |
Focus | Active Focus supports retries, alternative approaches and sustained execution |
Multimodality | Accepts text, images, video, audio and PDFs |
Coding | Trained for long-running projects and repeated test-and-revision cycles |
Availability | Direct Model API access for Wonder; Wonder Full remains subscription-only |
A newer model is not automatically the right migration target. Compare Muse Wonder 1.0 with Muse Spark 1.3 using the same prompts, tools, source material, effort controls and acceptance rubric. Record quality, latency, retries, review time and total cost per accepted result.
Capabilities and best use cases
Meta positions Muse Wonder for agents that operate across browsers and applications, produce artifacts and maintain plans over time. The model can delegate to subagents and use parallel tool calls through Meta's platform. These capabilities increase usefulness, but also increase the impact of stale memory, prompt injection and overbroad permissions.
Capability | Practical use |
|---|---|
Active Memory | Carry forward relevant preferences, constraints and prior context |
Active Focus | Continue through blockers, revise plans and ask for clarification |
Professional agents | Research, planning, finance, legal and health-information workflows |
Coding | Long-running projects, repository changes and iterative testing |
Multimodal input | Reason across text, images, video, audio and PDF files |
Subagent delegation | Parallelize bounded work through Meta's agent tooling |
The strongest documented fit is long-running personal and professional agents that need memory, sustained focus, coding and multimodal context. Reliability still depends on retrieval, tool permissions, prompt design, output validation and escalation policy. Test the complete system instead of evaluating the model name in isolation.
Benchmark evidence
Meta reports strong results across personal and professional agents, memory, coding and sustained execution. Its launch materials discuss Harvey's Legal Agent Benchmark, FocusBench and SWE-Marathon, among other evaluations. The public article does not provide every score in text, so this guide keeps the evidence boundary explicit rather than reconstructing values from charts.
Evaluation area | Reported evidence | How to interpret it |
|---|---|---|
Harvey Legal Agent Benchmark | Company reports frontier performance | Legal work products from source documents and detailed instructions |
FocusBench | Company reports strength | Measures sustained focus on challenging goals |
SWE-Marathon | Company reports strength | Long-running software projects across repeated cycles |
Active Memory evaluations | Company reports state-of-the-art results | Tests retrieval and use of earlier relevant context |
Out-of-distribution tests | Improves beyond the prior Spark trend | Meta analysis of generalization as test-time compute scales |
Benchmark results depend on the model version, effort, sampling, tools, scaffolding and dataset release. Do not combine scores from different harnesses into a synthetic ranking. Use public evidence to choose candidates, then test those candidates on frozen tasks from the intended workflow.
Pricing and access
Muse Wonder 1.0 is available through the Meta Model API and Muse Code. The API uses the identifier muse-wonder-1.0. Muse Wonder Full is a separate model available through eligible Muse Code subscriptions rather than the public token-priced API.
Meta lists a one-million-token context window with the same token rates throughout the window. The absence of a long-context surcharge does not make large prompts operationally free: latency, output growth, retrieval errors and tool usage still affect total cost per accepted result.
Usage or access item | Current value |
|---|---|
Standard input | $4 per million tokens |
Cached input | $0.20 per million tokens |
Standard output | $20 per million tokens |
Long-context premium | None documented |
Muse Wonder | Available in Meta Model API and Muse Code |
Muse Wonder Full | Available through eligible Muse Code subscriptions |
Token price is only part of deployment cost. Include prompt construction, cached-input behavior, tool calls, retries, wall time, failed-task recovery and human review. A cheaper model can be more expensive operationally when it needs repeated attempts or produces artifacts that require substantial correction.
Limitations and deployment risks
- Active Memory can preserve incorrect, outdated or overly sensitive information if memory controls are weak.
- Personal-agent access to messages, calendars and private files requires explicit permissions and auditability.
- Long-running agents can drift from the original objective or follow instructions embedded in untrusted content.
- Most launch evidence is provider-reported and should be reproduced on the intended workflow.
- Muse Wonder Full has a different access model, so its results should not be attributed automatically to Wonder 1.0.
- One-million-token prompts can create retrieval and latency problems even without a pricing multiplier.
- Important external actions need human confirmation and recoverable execution paths.
High-stakes legal, medical, financial, scientific and security work needs qualified review. Log the exact model version and settings, separate untrusted content from system instructions, minimize tool permissions and require confirmation before irreversible or externally visible actions.
How to evaluate Muse Wonder 1.0
Build a frozen evaluation set of 20 to 50 real tasks. Include routine work, difficult edge cases, adversarial inputs, long-context examples and cases where the correct behavior is to stop or escalate. Compare Muse Wonder 1.0 with Muse Spark 1.3 and at least one neighboring model under equivalent conditions.
Test area | What to record |
|---|---|
Task completion | Pass or fail against a written acceptance rubric |
Reliability | Repeated-run success rate, variance and silent failures |
Factuality | Unsupported claims, source use, quotations and citation accuracy |
Tool use | Wrong calls, retries, recovery and permission-boundary failures |
Long context | Retrieval accuracy, instruction retention and cost at realistic lengths |
Efficiency | Wall time, input, output, cached tokens, tool charges and review time |
Safety | Prompt injection, sensitive-data handling and irreversible-action controls |
Choose the least expensive configuration that clears the acceptance threshold with a safety margin. Re-run the suite when the model, system prompt, effort, retrieval layer, tool definitions or approval policy changes.
Who should use it?
Situation | Recommendation |
|---|---|
Strong fit | long-running personal and professional agents that need memory, sustained focus, coding and multimodal context |
Pilot first | Long-running agents, large contexts, computer use and workflows with several tools |
Escalate | Ambiguous or consequential work that does not reliably clear the evaluation threshold |
Avoid unsupervised use | Irreversible actions, sensitive data or high-stakes decisions without monitoring and approval |
A migration should be driven by measured outcomes. Keep Muse Spark 1.3 available during the pilot, record where each model succeeds or fails, and use routing when different task classes have different quality and cost requirements.
Frequently asked questions
Is Muse Wonder 1.0 open source?
No. Muse Wonder 1.0 is a proprietary model controlled by Meta. Access, lifecycle and serving behavior depend on the provider and supported distribution platforms.
How much does Muse Wonder 1.0 cost?
$4 per million tokens. Review the full table because caching, prompt length, processing mode, platform and output volume can materially change the total.
What is Muse Wonder 1.0 best used for?
Its strongest documented fit is long-running personal and professional agents that need memory, sustained focus, coding and multimodal context. Start with a supervised pilot and retain human sign-off for consequential work.
Should I migrate from Muse Spark 1.3?
Only after a side-by-side evaluation. Measure accepted-task quality, total cost, latency, output style, tool reliability and migration engineering. Newer does not always mean better for a specific operational workflow.
How should benchmark evidence for Muse Wonder 1.0 be interpreted?
Use provider results to identify promising workloads, then reproduce the comparison with the exact configuration, tools and acceptance criteria that matter to the deployment. Do not infer a site ranking from benchmark claims gathered under a different harness.
Related model guides
Browse the Best AI Models directory.
Compare with the Muse Spark 1.3 guide.
Compare with the Muse Spark 1.1 guide.
Compare with the Claude Opus 5.5 guide.
Official sources and update policy
Meta AI Muse Wonder announcement.
Meta Model API Muse Wonder page.
Meta Model API pricing and rate limits.
Checked October 22, 2026. We update this guide when Meta changes the specification, pricing, access, safety documentation or lifecycle. Vendor benchmark claims are attributed and are not presented as independent testing. No vendor payment or affiliate relationship determined inclusion.