Meta: Muse Wonder Full

Muse Wonder Full is the most capable model in Meta's Muse Wonder series, built for long-running professional agents and ultra-long-horizon software engineering. It shares Muse Wonder's Active Memory and Active Focus capabilities and generally scores higher on Meta's agentic and coding evaluations.

Model lab: Meta
Model family: muse

Model resources

Announcement Model card

Modalitiestext, image, video, audio → text

Context window1,048,576

Maximum output—

Try now

Quickstart

Use your Moduck API Key, set the API base URL to https://api.moduck.ai/v1, and set the model parameter to meta/muse-wonder-full.

import os
import requests

base_url = "https://api.moduck.ai/v1"
api_key = os.environ["MODUCK_API_KEY"]
model_id = "meta/muse-wonder-full"

response = requests.post(
    f"{base_url}/chat/completions",
    headers={
        "Authorization": f"Bearer {api_key}",
        "Content-Type": "application/json",
    },
    json={
        "model": model_id,
        "messages": [{"role": "user", "content": "Hello!"}],
        "stream": False,
    },
    timeout=60,
)
response.raise_for_status()
print(response.json())

Use the familiar OpenAI-style chat request format to make model integration easier.

Muse Wonder Full API Pricing

Price currently unavailable

Benchmarks

Max

Intelligence Index · v4.355.8

Coding Index—

Agentic Index—

Benchmark	Score	Measurement
Harvey Legal Agent Benchmark	13.1	all-pass rate · Variant: max effort; Artificial Analysis LAB-AA · Version: 1.1 · Date: 2026-10-12
Finance Agent	59.1	accuracy · Dataset: Vals AI private test set · Variant: max effort · Version: v2 · Date: 2026-10-12
HealthBench Professional	62.3	length-normalized weighted rubric score · Variant: max effort · Date: 2026-10-12
AutomationBench	45.4	strict pass@1 · Variant: max effort; Artificial Analysis AutomationBench-AA · Date: 2026-10-12
GDP.pdf	33.2	score · Variant: max effort · Date: 2026-10-12
JobBench	75.3	mean reward · Variant: max effort · Date: 2026-10-12
OSWorld	78.4	mean partial score · Variant: max effort · Version: 2.1 · Date: 2026-10-12
OSWorld	39.4	binary completion rate · Variant: max effort · Version: 2.1 · Date: 2026-10-12
Terminal-Bench	57.1	accuracy · Variant: max effort; Artificial Analysis Terminal-Bench 4.0-AA · Version: 4.0 · Date: 2026-10-12
DeepSWE	79	pass@1 · Variant: max effort · Version: 1.1 · Date: 2026-10-12
SWE-Marathon	57.5	pass@1 · Variant: xhigh effort · Date: 2026-10-12
ProgramBench	59	share of rollouts passing at least 95% of tests · Variant: max effort · Harness: mini-swe-agent · Date: 2026-10-12
Artificial Analysis Intelligence Index	56	index score · Variant: max effort · Version: 4.3 · Date: 2026-10-12

Performance

Measured on the Moduck gateway across live traffic, hour by hour over the last 24 hours. Time to first token is how long before anything appears; throughput is how fast the answer is written after that; end-to-end latency is the whole round trip. Figures are medians unless the card says otherwise.

Performance trends

Throughput93.1 tok/sCurrent

Time to first token1.3 sCurrent

End-to-end latency7.8 sCurrent

Throughput

Current93.1 tok/s

Window average89.2 tok/s

Time to first token

Current1.3 s

Window average1.4 s

End-to-end latency

Current7.8 s

Window average7.5 s

Tail latency (P99)

Current16.4 s

Window average16.5 s

Prompt cache hit rate

Current18.1%

Window average24.6%

Error rate

Current0.0%

Window average0.1%

Availability

Uptime is the share of time at least one provider serving this model responded. Availability is stricter: the share of requests actually served. A provider that answers but fails part of the traffic lowers availability without lowering uptime. Both cover the last three days.

Uptime100%At least one provider responded

Availability100%Requests actually served

Error rate0%Share of failed requests

Served requests, hour by hour

24 hours agoNow

Muse Wonder Full Compare models

Model	Input price	Output price	Context window	Capabilities
Muse Wonder FullThis page	Price currently unavailable	Price currently unavailable	1,048,576	Text output · Reasoning · Tool calling
Muse Spark Open	USD 1 / 1M tokens	USD 3.4 / 1M tokens	1,048,576	Text output · Reasoning · Tool calling
Muse Wonder	USD 3.2 / 1M tokens	USD 16 / 1M tokens	1,048,576	Text output · Reasoning · Tool calling
Muse Spark 1.3	USD 1 / 1M tokens	USD 3.4 / 1M tokens	1,048,576	Text output · Reasoning · Tool calling
Muse Glimmer 30B	USD 0.2 / 1M tokens	USD 0.8 / 1M tokens	131,072	Text output · Reasoning · Tool calling

Model specifications

Input: text, image, video, audio
Output: text
Context window: 1,048,576
Maximum output: —
Release date: 2026-10-12
Attachments: Yes
Reasoning: Yes
Tool calling: Yes
Open weights: No

Frequently asked questions

What is Muse Wonder Full for?

What inputs and outputs does Muse Wonder Full support?

Recorded inputs: text, images, video, audio; outputs: text.

How is Muse Wonder Full priced?

This page does not list a price for this model yet.

What specifications are recorded for Muse Wonder Full?

Current model records list Context window 1,048,576 tokens.

What model parameter does the Quickstart use?

The example uses meta/muse-wonder-full. This value comes from the page identifier and is not confirmed as a callable Moduck public ID.

Can I call this model through the Moduck API now?

This page has no verified Moduck API call for this model. A Quickstart example or reserved ID does not establish access; check the models available to your account.