Long-horizon coding
78.8 on DeepSWE 1.1 and 53.1 on SWE-Marathon in Meta's results, ahead of Claude Opus 5.5, GPT-6 Astra and Claude Fable 5.1 on both.
Meta
Operational· checked every few minutesMuse Wonder API
The Muse Wonder API, Meta's first frontier model, for coding, long-running agents and computer use, 70% below Meta's list price
Price per 1M tokens70%off
- Input
- $1.20
$4.00 - Output
- $6.00
$20.00 - Cached input
- $0.06
Model ID
muse-wonder-1.0chat.py
from openai import OpenAIclient = OpenAI( base_url="https://api.zurelay.com/v1", api_key="YOUR_ZURELAY_KEY",)stream = client.chat.completions.create( model="muse-wonder-1.0", messages=[{"role": "user", "content": "Explain quicksort in two sentences."}], stream=True,)for chunk in stream: if chunk.choices: print(chunk.choices[0].delta.content or "", end="")Base URL https://api.zurelay.com/v1. Works with any OpenAI SDK. Chat API docs
- Context window
- 1M tokens
- Reasoning
- minimal to max
- Input
- Text, Image, PDF, Audio, Video
- Released
- Oct 12, 2026
- Status
- Operational
Pricing
Muse Wonder API pricing. A fraction of the list price.
Pay as you go from prepaid credit, with no subscription. Top up from $10. Requests that fail are never billed.
| Rate | Zurelay | Meta list | You save |
|---|---|---|---|
Input per 1M tokens | $1.20 | 70%off | |
Output per 1M tokens | $6.00 | 70%off | |
Cached input per 1M tokens | $0.06 | — | — |
One rate at every prompt length, up to the full 1M context window. Streaming, tool calls and structured outputs cost nothing extra.
Savings calculator
Move the sliders to your monthly usage.
Input tokens per month50M
Output tokens per month10M
Zurelay
$120.00
Meta list price
$400.00
You save every month$280.00
$3,360 a year
Overview
What is Muse Wonder?
Muse Wonder is the frontier model Meta Superintelligence Labs released on October 12, 2026: a much larger class of model than the Muse Spark family, built for coding in large repositories, agents that run for hours and computer use. The Zurelay Muse Wonder API serves it through one OpenAI-compatible endpoint with a 1M-token context window, at 70% below Meta's list price.
Meta Superintelligence Labs released Muse Wonder on October 12, 2026, as muse-wonder-1.0 in the Meta Model API, where Meta labels it a preview. It is Meta's first frontier model and a closed one, unlike the open Llama models. It reads text, images, video, audio and PDFs and writes text, with a context window of 1,048,576 tokens. On Zurelay the model ID is muse-wonder-1.0, and muse-wonder and meta/muse-wonder work too. It costs $1.20 per 1M input tokens and $6.00 per 1M output tokens.
Meta built Wonder for demanding work: making and checking changes across many files in large repositories, agents that plan long tasks, hand parts to sub-agents, call tools in parallel and recover from errors, and agents that operate browsers and desktop and mobile apps. Meta calls two of its traits active memory, which carries important context through a conversation, and active focus, which keeps the model working toward a goal when a task gets hard.
In Meta's own results at maximum reasoning, Muse Wonder scores 78.8 on DeepSWE 1.1, ahead of Claude Opus 5.5 (74.2), GPT-6 Astra (74.1) and Claude Fable 5.1 (67.4), and 53.1 on SWE-Marathon, ahead of all three. Its 85.6 on MemoryBench is the highest in Meta's table. It trails Opus 5.5 and GPT-6 Astra on Terminal-Bench 4.0 (52.0 against 59.6 and 59.1), and Artificial Analysis puts it at 52 on its Intelligence Index.
Muse Wonder thinks before it answers, at a level from minimal to max. Meta's price is $4.00 input, $20.00 output and $0.20 cached input per 1M tokens, the same across the whole context window. A larger version, Muse Wonder Full, is only in Meta's Muse Code subscriptions and has no API; asking Wonder for max reasoning doesn't turn it into Full.
Strengths
Where Muse Wonder shines. And what teams build with it.
Memory across a conversation
85.6 on MemoryBench, the top score in Meta's table, for keeping track of what matters over a long exchange.
Agents that run for hours
Plans long tasks, delegates parts to sub-agents, calls tools in parallel and recovers from errors, as Meta describes it.
Every input type
Reads text, images, PDFs, audio and video in one 1,048,576-token context, and answers in text.
Use cases
Coding agents
Changes across many files in large repositories, made and checked in parallel, in agents that run tools over many steps.
Computer use
Agents that see and operate browsers and desktop and mobile apps to finish everyday tasks.
Assistants with memory
Long conversations and personal assistants that need to remember and act on what a person said earlier.
Mixed-media analysis
Recordings, videos, screenshots and long documents read together in one request.
Get started
Call Muse Wonder in three steps.
No waitlist and no new SDK. If your code already talks to OpenAI, it already talks to Zurelay.
- 1
Create a key
Sign up, add credit and create an API key. Set a monthly budget or a rate limit per key if you like.
- 2
Point your SDK at Zurelay
Change the base URL. Everything else in your code stays the same.
https://api.zurelay.com/v1 - 3
Use muse-wonder-1.0
Send chat completions as usual. Streaming, tool calls and usage reporting work as you expect.
muse-wonder-1.0
Works with the tools you already use
Compare
Similar models. All on the same key.
65%off
Claude Opus 5.5
$1.40 / $7.00 per 1M tokens
Pick Claude Opus 5.5 to compare Anthropic's Opus model, which leads Wonder on Terminal-Bench 4.0 in Meta's own table.
Claude Opus 5.5 API75%off
GPT-6 Astra
$2.50 / $12.50 per 1M tokens
Pick GPT-6 Astra to compare OpenAI's frontier model on agentic coding and professional work.
GPT-6 Astra API60%off
Gemini 4 Argon
$0.80 / $4.00 per 1M tokens
Pick Gemini 4 Argon for Google's frontier model, which also reads text, images, PDFs, audio and video, at a lower price per token.
Gemini 4 Argon APIHow much does the Muse Wonder API cost?
On Zurelay, muse-wonder-1.0 costs $1.20 per 1M input tokens and $6.00 per 1M output tokens, with cached input at $0.06. Meta's price is $4.00 input and $20.00 output per 1M, so you save 70%. The rate is the same across the whole context window, and reasoning tokens bill as output.
What is Muse Wonder best at?
Long, demanding work: coding across large repositories, agents that run for hours and call tools in parallel, computer use, and assistants that need to remember context. Meta reports 78.8 on DeepSWE 1.1, 53.1 on SWE-Marathon and 85.6 on MemoryBench at maximum reasoning.
What is the difference between Muse Wonder and Muse Wonder Full?
Muse Wonder Full is a larger version that Meta offers only inside its Muse Code subscriptions, with no API, so no API provider can sell it. Muse Wonder is the model in Meta's API and the one Zurelay serves. Setting reasoning_effort to max makes Wonder think longer; it doesn't turn it into Full.
How do I call the Muse Wonder API with the OpenAI SDK?
Install the official openai package, set base_url to https://api.zurelay.com/v1 and use your Zurelay API key. Then send a Chat Completions request with model set to "muse-wonder-1.0". Images, PDFs, audio and video go in the message as data or links.
What is the Muse Wonder context window?
Muse Wonder has a 1,048,576-token context window. It accepts text, image, PDF, audio and video input and writes text.
How do Muse Wonder reasoning levels work?
Set reasoning_effort to minimal, low, medium, high, xhigh or max; none isn't offered, and a request for none runs at minimal. Without a level the model picks one itself, which in our tests thinks about as long as high. Muse Wonder sends nothing while it thinks, and a request that thinks for more than about 45 seconds before its first word can time out; failed requests aren't charged, so for the hardest problems try high before xhigh or max.
Can I force Muse Wonder to call a tool?
Yes. Muse Wonder calls tools you define, several at once if it needs to, and Meta's API only takes tool_choice "auto". So when you send "required" or a named function, Zurelay sends auto, offers only the tool you named and tells the model to call it; for "none", it tells the model to answer in text.
Is it the same model as Meta's Muse Wonder?
Yes, it is Meta's Muse Wonder, muse-wonder-1.0. Responses are sampled, so wording varies between runs, and Zurelay does not promise byte-identical output. Zurelay is independent and is not affiliated with or endorsed by Meta.
More models on Zurelay
All models- Zurelay Auto APIfrom $0.025 / $0.12 per 1M tokensAuto
- Claude Opus 5.5 API$1.40 / $7.00 per 1M tokens−65%
- GPT-6 Astra API$2.50 / $12.50 per 1M tokens−75%
- Nano Banana 2.1 APIfrom $0.03 per image−11%
- Nano Banana Pro APIfrom $0.035 per image−74%
- Nano Banana 2 APIfrom $0.025 per image−63%
- Nano Banana 2 Lite API$0.016 per image−52%
- Claude Sonnet 5.5 API$0.90 / $4.50 per 1M tokens−55%
- Claude Haiku 5.5 API$0.04 / $0.20 per 1M tokens−60%
- GPT-6.1 Sol API$0.50 / $2.50 per 1M tokens−75%
- GPT-6 Sol API$0.50 / $2.50 per 1M tokens−75%
- GPT-6 Luna API$0.025 / $0.125 per 1M tokens−75%
- Grok 4.7 API$0.70 / $2.10 per 1M tokens−65%
- MiMo V2.6 Pro API$0.32 / $0.64 per 1M tokens−26%
- MiMo V2.6 Flash API$0.075 / $0.15 per 1M tokens−46%
- DeepSeek V4.1 Flash 1017 API$0.03 / $0.12 per 1M tokens−90%
- DeepSeek V4.1 Flash API$0.03 / $0.12 per 1M tokens−90%
- DeepSeek V4 Flash API$0.03 / $0.12 per 1M tokens−90%
- DeepSeek V4 Flash 0731 API$0.029 / $0.087 per 1M tokens−43%
- DeepSeek V4 Pro 0813 API$0.23 / $0.62 per 1M tokens−84%
- Claude Fable 5.1 API$3.50 / $17.50 per 1M tokens−65%
- Claude Fable 5 API$3.50 / $17.50 per 1M tokens−65%
- Claude Opus 5 API$1.75 / $8.75 per 1M tokens−65%
- Claude Opus 4.8 API$2.40 / $12.00 per 1M tokens−52%
- Claude Opus 4.7 API$1.78 / $8.93 per 1M tokens−64%
- Claude Opus 4.6 API$2.40 / $12.00 per 1M tokens−52%
- Claude Haiku 4.5 API$0.54 / $2.70 per 1M tokens−46%
- GPT-5.6 Sol API$1.14 / $5.70 per 1M tokens−72%
- GPT-5.6 Terra API$0.54 / $3.26 per 1M tokens−73%
- GPT-5.6 Luna API$0.08 / $0.49 per 1M tokens−59%
- Gemini 4 Argon API$0.80 / $4.00 per 1M tokens−60%
- Gemini 3.1 Pro API$0.80 / $4.80 per 1M tokens−60%
- Gemini 3.8 Flash API$0.32 / $1.59 per 1M tokens−58%
- Gemini 3.7 Flash API$0.32 / $1.59 per 1M tokens−58%
- Gemini 3.5 Flash-Lite API$0.12 / $1.04 per 1M tokens−58%
- Gemini 3.1 Flash-Lite API$0.10 / $0.59 per 1M tokens−61%
- GLM-5.3 API$0.38 / $1.21 per 1M tokens−73%
- GLM-5.2 API$0.38 / $1.21 per 1M tokens−73%
- GLM-5.3 Flash API$0.07 / $0.25 per 1M tokens−50%
- Kimi K3.1 Pro API$0.80 / $4.00 per 1M tokens−60%
- Kimi K3.1 Pro Highspeed API$3.30 / $16.50 per 1M tokens−18%
- Kimi K3 API$1.27 / $6.37 per 1M tokens−58%
- Qwen 3.8 Max API$1.20 / $3.60 per 1M tokens−40%
- Qwen 3.8 Flash API$0.07 / $0.22 per 1M tokens−53%
- Qwen 3.8 Omni Flash API$0.075 / $0.235 per 1M tokens−50%
- MiniMax M3.1 Flash API$0.06 / $0.24 per 1M tokens−60%
- HY4 Preview API$0.40 / $1.20 per 1M tokens−52%
- MiniMax M3 API$0.13 / $0.51 per 1M tokens−57%
- MiniMax M2.7 API$0.13 / $0.51 per 1M tokens−57%
- GPT Image 2.5 Flare APIfrom $0.028 per image−47%
- GPT Image 2.5 Sunburst APIfrom $0.026 per image−51%
- GPT Image 2 API$0.02 per image−62%
- Grok Imagine Image 2.0 APIfrom $0.045 per image−25%
- FLUX 3 Image APIfrom $0.04 per image−17%
- Seedance 2.5 APIfrom $0.135 per secondVideo
- Seedance 2.0 APIfrom $0.097 per secondVideo
- Seedance 2.0 Fast APIfrom $0.074 per secondVideo
- Seedance 2.0 Mini APIfrom $0.047 per secondVideo
- Gemini Omni Flash APIfrom $0.042 per secondVideo
- Kling 3.0 APIfrom $0.125 per secondVideo
- Wan 3.0 APIfrom $0.05 per secondVideo
- Veo 3.1 Fast APIfrom $0.025 per secondVideo
- Grok Imagine Video 1.5 APIfrom $0.071 per secondVideo
- Vidu Q3 Turbo APIfrom $0.03 per secondVideo
- PixVerse V6 APIfrom $0.021 per secondVideo
- MiniMax H3 APIfrom $0.065 per secondVideo
- Gemini 3.8 Flash TTS API$0.30 / $5.40 per 1M tokens−40%
- Gemini 3.1 Flash TTS API$0.60 / $12.00 per 1M tokens−40%
- Gemini 2.5 Pro TTS API$0.60 / $12.00 per 1M tokens−40%
- Gemini 2.5 Flash TTS API$0.30 / $6.00 per 1M tokens−40%
- Jev 1.14 API$0.024 per 1M input tokens−43%
- Jev 1.13 API$0.035 per 1M input tokens−17%
Stop paying list price.
Start saving today.
Create a key in seconds, point the SDK you already use at Zurelay, and every request costs up to 90% less from the first token.