Muse Wonder API
Meta's first frontier model family, built for long-running personal and professional agent work, served by Meta directly.
Model spec
- Input
- $4.00per 1M tokens
- Output
- $20.00per 1M tokens
- Cached input
- $0.20per 1M tokens
- Context
- 1.05M
Model ID
OpenAI-compatible · api.companyfabric.com/v1
Overview
About Muse Wonder
Muse Wonder is Meta's first frontier model family, released by Meta Superintelligence Labs on 12 October 2026 for long-running personal and professional agent work: research, finance and legal analysis, business automation and long-horizon coding. It has a context window of about 1M tokens, reads images, and reasons before answering. Meta serves it through its own Meta Model API at $4 input and $20 output per million tokens, with cached input at $0.20, and CompanyFabric routes it to Meta directly, on the same key and balance as every other model. The bigger Muse Wonder Full is not sold through the API; Meta keeps it inside its Muse Code subscription. Every model on CompanyFabric shares the same API key, billing balance, and rate-limit envelope, one integration covers the entire library.
Facts
At a glance
Availability
Live
Added 12 Oct 2026
Pricing
$4.00 in · $20.00 out
Per 1M tokens, metered exactly per call
Context window
1.05M tokens
Commercial use
See license
Commercial output use permitted per Meta Model API terms. Closed weights, API only.
Variants
All endpoints
Muse Spark 1.3
$1.25/Mtok inMeta Superintelligence Labs' frontier model, fourth on the Artificial Analysis index, served by Meta directly.
Chat Completions
Muse Wonder
$4/Mtok inMeta's first frontier model family, built for long-running personal and professional agent work, served by Meta directly.
Chat Completions
Quickstart
How to use the Muse Wonder API
- 1
Create a free CompanyFabric account, starter credits are included, no card required.
- 2
Copy your API key from the dashboard. One key covers every model in the library.
- 3
Point any OpenAI SDK at api.companyfabric.com/v1 and set model to "meta/muse-wonder".
- 4
Read the exact metered cost of the call from the response, the same number shown in the playground.
Use cases
What teams build with it
Agents & copilots
Drive agent loops, tool calls, and user-facing assistants over the OpenAI-compatible endpoint.
agentstool useassistants
Content pipelines
Summarization, rewriting, and structured generation at metered per-token cost.
contentsummarization
Extraction & classification
Turn unstructured input into typed JSON with schema-constrained output.
extractionjson mode
Get better output
Prompting tips
- State the output format before the task, models follow contracts declared early.
- Few-shot beats description: two worked examples outperform a paragraph of rules.
- Cap output length explicitly when cost matters; output tokens dominate most bills.
Exact, metered, prepaid
Muse Wonder pricing
Muse Wonder pricing is per token: $4 / M input · $20 / M output · $0.2 / M cached input. Output tokens cost 5× input tokens, so long generations, not long prompts, are what actually drive most bills. A typical call with a 2k-token prompt and a 1k-token answer costs about $0.0280.
Every call is metered exactly and attributed to your key, the dashboard, the API usage response, and your invoice all read from the same ledger row. Cap any key's monthly spend with a budget cap, and bring your own provider key (BYOK) at 0% platform fee if you already have a direct contract.
| Endpoint | Type | Price |
|---|---|---|
| Muse Spark 1.3 | Chat | $1.25 / M input · $4.25 / M output |
| Muse Wonder | Chat | $4 / M input · $20 / M output · $0.2 / M cached input |
Starter credits
Use Muse Wonder free
Your CompanyFabric starter credits cover your first Muse Wonder calls at no cost, create a free account, no card required, and the credits apply to Muse Wonder and every other model on the key. When they run out, prepaid credits or BYOK (0% fee) take over with no plan to pick.
Production-grade
Code samples
import OpenAI from "openai"
const client = new OpenAI({
baseURL: "https://api.companyfabric.com/v1",
apiKey: process.env.COMPANYFABRIC_API_KEY,
})
const res = await client.chat.completions.create({
model: "meta/muse-wonder",
max_tokens: 1024,
temperature: 0.7,
messages: [{ role: "user", content: "Say hello from the fabric." }],
})
console.log(res.choices[0].message.content)Playground
Try it in the browser
Comparisons
Muse Wonder vs the alternatives
Muse Wonder vs Claude Fable 5.1
Muse Wonder and Claude Fable 5.1 compete for the same text workloads. Both run behind the same CompanyFabric key at exact metered prices, A/B them on your real prompts and let the outputs decide, then switch with a one-line change.
Full comparison: Muse Wonder vs Claude Fable 5.1 →Muse Wonder vs GPT-6 Astra
Muse Wonder and GPT-6 Astra compete for the same text workloads. Both run behind the same CompanyFabric key at exact metered prices, A/B them on your real prompts and let the outputs decide, then switch with a one-line change.
Full comparison: Muse Wonder vs GPT-6 Astra →Muse Wonder vs Claude Opus 5.5
Muse Wonder and Claude Opus 5.5 compete for the same text workloads. Both run behind the same CompanyFabric key at exact metered prices, A/B them on your real prompts and let the outputs decide, then switch with a one-line change.
Full comparison: Muse Wonder vs Claude Opus 5.5 →Compare
Muse Wonder head to head
FAQ
Frequently asked questions
How do I get a Muse Wonder API key?
Create a CompanyFabric account, and one key unlocks Muse Wonder and every other model in the library. Free starter credits are included, no subscription, no card required to try it.
Why use Muse Wonder on CompanyFabric instead of going direct?
One key, one prepaid balance, and one rate-limit envelope across the whole library, at pricing at parity with or below the direct API. Model switching is a one-line change, and BYOK routing is 0% fee.
Is the API OpenAI-compatible?
Yes, point any OpenAI SDK at api.companyfabric.com/v1 and set model to "meta/muse-wonder".
Can I use the output commercially?
Commercial output use permitted per Meta Model API terms. Closed weights, API only.
How does billing work?
Prepaid credits via card, or BYOK with your own provider keys at 0% platform fee. No subscription required; balances never expire.
How much does the Muse Wonder API cost?
$4 per million input tokens, $20 per million output tokens, and $0.20 per million cached input tokens, the Meta Model API's launch price. Long agent runs resend the same context on every step, so most of their input is cached and bills at the cached rate. Each response carries its exact cost in usage.cost_usd.
What is the difference between Muse Wonder and Muse Wonder Full?
Muse Wonder is the model Meta sells through its API, and the one CompanyFabric serves. Muse Wonder Full is the larger, slower and stronger version that tops Meta's own table; at launch it is available only inside Meta's Muse Code subscription, with no per-token API. If Meta opens it to the API, it will be added here.
How does Muse Wonder compare with Claude Fable 5.1, GPT-6 Astra and Claude Opus 5.5?
On Meta's launch table, all at maximum effort, Muse Wonder leads the three on Harvey LAB-AA legal agent work (10.0 against 6.4, 8.6 and 4.2), MemoryBench (85.6), DeepSWE (78.8) and SWE-Marathon (53.1). It trails Claude Opus 5.5 and GPT-6 Astra on Terminal-Bench 4.0 (52.0 against 59.6 and 59.1) and scores 52 on the Artificial Analysis Index, against 58 for Opus 5.5. These are Meta's numbers, not independent results.
Is Muse Wonder used by companyfabric/auto?
No. auto, cheap and fast pick the cheapest model that fits the request, and best is unchanged, so Muse Wonder is only used when you name it: "model": "meta/muse-wonder".
Provider
About Meta
Meta publishes the muse family. Also in this family: Muse Spark 1.3. License: Commercial output use permitted per Meta Model API terms. Closed weights, API only.
One key. Every model. Exact prices.
Run Muse Wonder with free starter credits. Free starter credits included, no subscription required.